Remove Compute Vision CL support

Resolves COMPMID-4151

Change-Id: I46f541efe8c4087f27794d2e158b6c1547d459ba
Signed-off-by: Michalis Spyrou <michalis.spyrou@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5160
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>

68 files changed, 0 insertions, 9145 deletions

diff --git a/src/core/CL/kernels/CLAbsoluteDifferenceKernel.cpp b/src/core/CL/kernels/CLAbsoluteDifferenceKernel.cpp
deleted file mode 100644
index 76b60cb9f8..0000000000
--- a/src/core/CL/kernels/CLAbsoluteDifferenceKernel.cpp
+++ /dev/null
@@ -1,104 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Validate.h"
-
-#include "src/core/CL/kernels/CLAbsoluteDifferenceKernel.h"
-#include "src/core/helpers/WindowHelpers.h"
-
-#include <set>
-#include <string>
-
-using namespace arm_compute;
-
-CLAbsoluteDifferenceKernel::CLAbsoluteDifferenceKernel()
-    : _input1(nullptr), _input2(nullptr), _output(nullptr)
-{
-}
-
-void CLAbsoluteDifferenceKernel::configure(const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input1, input2, output);
-}
-
-void CLAbsoluteDifferenceKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input1, 1, DataType::U8, DataType::S16);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input2, 1, DataType::U8, DataType::S16);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8, DataType::S16);
-    ARM_COMPUTE_ERROR_ON_MSG(output->info()->data_type() == DataType::U8 && (input1->info()->data_type() != DataType::U8 || input2->info()->data_type() != DataType::U8),
-                             "The output image can only be U8 if both input images are U8");
-
-    _input1 = input1;
-    _input2 = input2;
-    _output = output;
-
-    // Set kernel build options
-    std::set<std::string> build_opts;
-    build_opts.insert("-DDATA_TYPE_IN1=" + get_cl_type_from_data_type(input1->info()->data_type()));
-    build_opts.insert("-DDATA_TYPE_IN2=" + get_cl_type_from_data_type(input2->info()->data_type()));
-    build_opts.insert("-DDATA_TYPE_OUT=" + get_cl_type_from_data_type(output->info()->data_type()));
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "absdiff", build_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 16;
-
-    Window win = calculate_max_window(*input1->info(), Steps(num_elems_processed_per_iteration));
-
-    AccessWindowHorizontal input1_access(input1->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal input2_access(input2->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win, input1_access, input2_access, output_access);
-
-    ValidRegion valid_region = intersect_valid_regions(input1->info()->valid_region(),
-                                                       input2->info()->valid_region());
-
-    output_access.set_valid_region(win, valid_region);
-
-    ICLKernel::configure_internal(win);
-}
-
-void CLAbsoluteDifferenceKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input1, slice);
-        add_2D_tensor_argument(idx, _input2, slice);
-        add_2D_tensor_argument(idx, _output, slice);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLAbsoluteDifferenceKernel.h b/src/core/CL/kernels/CLAbsoluteDifferenceKernel.h
deleted file mode 100644
index 28f28fe44f..0000000000
--- a/src/core/CL/kernels/CLAbsoluteDifferenceKernel.h
+++ /dev/null
@@ -1,79 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLABSOLUTEDIFFERENCEKERNEL_H
-#define ARM_COMPUTE_CLABSOLUTEDIFFERENCEKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the absolute difference kernel.
- *
- * Absolute difference is computed by:
- * @f[ output(x,y) = | input1(x,y) - input2(x,y) | @f]
- */
-class CLAbsoluteDifferenceKernel : public ICLKernel
-{
-public:
-    /** Default constructor. */
-    CLAbsoluteDifferenceKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLAbsoluteDifferenceKernel(const CLAbsoluteDifferenceKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLAbsoluteDifferenceKernel &operator=(const CLAbsoluteDifferenceKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLAbsoluteDifferenceKernel(CLAbsoluteDifferenceKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLAbsoluteDifferenceKernel &operator=(CLAbsoluteDifferenceKernel &&) = default;
-    /** Default destructor */
-    ~CLAbsoluteDifferenceKernel() = default;
-
-    /** Set the inputs and output images.
-     *
-     * @param[in]  input1 Source tensor. Data types supported: U8/S16.
-     * @param[in]  input2 Source tensor. Data types supported: U8/S16.
-     * @param[out] output Destination tensor. Data types supported: U8/S16.
-     */
-    void configure(const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output);
-    /** Set the inputs and output images.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input1          Source tensor. Data types supported: U8/S16.
-     * @param[in]  input2          Source tensor. Data types supported: U8/S16.
-     * @param[out] output          Destination tensor. Data types supported: U8/S16.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_input1; /**< Source tensor 1. */
-    const ICLTensor *_input2; /**< Source tensor 2. */
-    ICLTensor       *_output; /**< Destination tensor. */
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLABSOLUTEDIFFERENCEKERNEL_H */
diff --git a/src/core/CL/kernels/CLAccumulateKernel.cpp b/src/core/CL/kernels/CLAccumulateKernel.cpp
deleted file mode 100644
index b0a8eba644..0000000000
--- a/src/core/CL/kernels/CLAccumulateKernel.cpp
+++ /dev/null
@@ -1,101 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLAccumulateKernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Error.h"
-#include "arm_compute/core/IAccessWindow.h"
-#include "arm_compute/core/Types.h"
-#include "arm_compute/core/Validate.h"
-
-namespace arm_compute
-{
-namespace
-{
-constexpr unsigned int num_elems_processed_per_iteration = 16;
-} // namespace
-
-void CLAccumulateKernel::configure(const ICLTensor *input, ICLTensor *accum)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, accum);
-}
-
-void CLAccumulateKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *accum)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(accum, 1, DataType::S16);
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "accumulate");
-
-    // Make sure _kernel is initialized before calling the parent's configure
-    ICLSimple2DKernel::configure(input, accum, num_elems_processed_per_iteration);
-}
-
-void CLAccumulateWeightedKernel::configure(const ICLTensor *input, float alpha, ICLTensor *accum)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, alpha, accum);
-}
-
-void CLAccumulateWeightedKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, float alpha, ICLTensor *accum)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(accum, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON(alpha < 0.0 || alpha > 1.0);
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "accumulate_weighted");
-
-    // Set static kernel arguments
-    unsigned int idx = 2 * num_arguments_per_2D_tensor(); //Skip the input and output parameters
-    _kernel.setArg(idx++, alpha);
-
-    // Configure kernel window
-    ICLSimple2DKernel::configure(input, accum, num_elems_processed_per_iteration);
-}
-
-void CLAccumulateSquaredKernel::configure(const ICLTensor *input, uint32_t shift, ICLTensor *accum)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, shift, accum);
-}
-
-void CLAccumulateSquaredKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, uint32_t shift, ICLTensor *accum)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(accum, 1, DataType::S16);
-    ARM_COMPUTE_ERROR_ON(shift > 15);
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "accumulate_squared");
-
-    // Set static kernel arguments
-    unsigned int idx = 2 * num_arguments_per_2D_tensor(); //Skip the input and output parameters
-    _kernel.setArg(idx++, shift);
-
-    // Configure kernel window
-    ICLSimple2DKernel::configure(input, accum, num_elems_processed_per_iteration);
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLAccumulateKernel.h b/src/core/CL/kernels/CLAccumulateKernel.h
deleted file mode 100644
index 16a715319d..0000000000
--- a/src/core/CL/kernels/CLAccumulateKernel.h
+++ /dev/null
@@ -1,114 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLACCUMULATEKERNEL_H
-#define ARM_COMPUTE_CLACCUMULATEKERNEL_H
-
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-#include <cstdint>
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the accumulate kernel.
- *
- * Accumulation is computed by:
- * @f[ accum(x,y) = accum(x,y) + input(x,y) @f]
- */
-class CLAccumulateKernel : public ICLSimple2DKernel
-{
-public:
-    /** Set the input and accumulation tensors.
-     *
-     * @param[in]  input Source tensor. Data types supported: U8.
-     * @param[out] accum Destination tensor. Data types supported: S16.
-     */
-    void configure(const ICLTensor *input, ICLTensor *accum);
-    /** Set the input and accumulation tensors.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Source tensor. Data types supported: U8.
-     * @param[out] accum           Destination tensor. Data types supported: S16.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *accum);
-};
-
-/** Interface for the accumulate weighted kernel.
- *
- * Weighted accumulation is computed:
- * @f[ accum(x,y) = (1 - \alpha)*accum(x,y) + \alpha*input(x,y) @f]
- *
- * Where @f$ 0 \le \alpha \le 1 @f$
- * Conceptually, the rounding for this is defined as:
- * @f[ output(x,y)= uint8( (1 - \alpha) * float32( int32( output(x,y) ) ) + \alpha * float32( int32( input(x,y) ) ) ) @f]
-*/
-class CLAccumulateWeightedKernel : public ICLSimple2DKernel
-{
-public:
-    /** Set the input and accumulation images, and the scale value.
-     *
-     * @param[in]     input Source tensor. Data types supported: U8.
-     * @param[in]     alpha Scalar value in the range [0, 1.0]. Data types supported: F32.
-     * @param[in,out] accum Accumulated tensor. Data types supported: U8.
-     */
-    void configure(const ICLTensor *input, float alpha, ICLTensor *accum);
-    /** Set the input and accumulation images, and the scale value.
-     *
-     * @param[in]     compile_context The compile context to be used.
-     * @param[in]     input           Source tensor. Data types supported: U8.
-     * @param[in]     alpha           Scalar value in the range [0, 1.0]. Data types supported: F32.
-     * @param[in,out] accum           Accumulated tensor. Data types supported: U8.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, float alpha, ICLTensor *accum);
-};
-
-/** Interface for the accumulate squared kernel.
- *
- * The accumulation of squares is computed:
- * @f[ accum(x,y) = saturate_{int16} ( (uint16) accum(x,y) + (((uint16)(input(x,y)^2)) >> (shift)) ) @f]
- *
- * Where @f$ 0 \le shift \le 15 @f$
-*/
-class CLAccumulateSquaredKernel : public ICLSimple2DKernel
-{
-public:
-    /** Set the input and accumulation tensors and the shift value.
-     *
-     * @param[in]     input Source tensor. Data types supported: U8.
-     * @param[in]     shift Shift value in the range of [0, 15]. Data types supported: U32.
-     * @param[in,out] accum Accumulated tensor. Data types supported: S16.
-     */
-    void configure(const ICLTensor *input, uint32_t shift, ICLTensor *accum);
-    /** Set the input and accumulation tensors and the shift value.
-     *
-     * @param[in]     compile_context The compile context to be used.
-     * @param[in]     input           Source tensor. Data types supported: U8.
-     * @param[in]     shift           Shift value in the range of [0, 15]. Data types supported: U32.
-     * @param[in,out] accum           Accumulated tensor. Data types supported: S16.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, uint32_t shift, ICLTensor *accum);
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLACCUMULATEKERNEL_H */
diff --git a/src/core/CL/kernels/CLBox3x3Kernel.cpp b/src/core/CL/kernels/CLBox3x3Kernel.cpp
deleted file mode 100644
index 9f493b4fb8..0000000000
--- a/src/core/CL/kernels/CLBox3x3Kernel.cpp
+++ /dev/null
@@ -1,81 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLBox3x3Kernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-
-#include <set>
-#include <string>
-
-using namespace arm_compute;
-
-BorderSize CLBox3x3Kernel::border_size() const
-{
-    return BorderSize(1);
-}
-
-void CLBox3x3Kernel::configure(const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, border_undefined);
-}
-
-void CLBox3x3Kernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8);
-
-    _input  = input;
-    _output = output;
-
-    // Set build options
-    std::set<std::string> build_opts = { "-DMAT0=1", "-DMAT1=1", "-DMAT2=1",
-                                         "-DMAT3=1", "-DMAT4=1", "-DMAT5=1",
-                                         "-DMAT6=1", "-DMAT7=1", "-DMAT8=1",
-                                         "-DSCALE=9", "-DDATA_TYPE_OUT=uchar"
-                                       };
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "convolution3x3_static", build_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-    constexpr unsigned int num_rows_read_per_iteration       = 3;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-}
diff --git a/src/core/CL/kernels/CLBox3x3Kernel.h b/src/core/CL/kernels/CLBox3x3Kernel.h
deleted file mode 100644
index 2373c4a928..0000000000
--- a/src/core/CL/kernels/CLBox3x3Kernel.h
+++ /dev/null
@@ -1,59 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLBOX3X3KERNEL_H
-#define ARM_COMPUTE_CLBOX3X3KERNEL_H
-
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the box 3x3 filter kernel.
- *
- */
-class CLBox3x3Kernel : public ICLSimple2DKernel
-{
-public:
-    /**Initialise the kernel's input and output.
-     *
-     * @param[in]  input            An input tensor. Data types supported: U8
-     * @param[out] output           The output tensor. Data types supported: U8.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, bool border_undefined);
-    /**Initialise the kernel's input and output.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            An input tensor. Data types supported: U8
-     * @param[out] output           The output tensor. Data types supported: U8.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined);
-
-    //Inherited methods overriden:
-    BorderSize border_size() const override;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLBOX3X3KERNEL_H */
diff --git a/src/core/CL/kernels/CLCannyEdgeKernel.cpp b/src/core/CL/kernels/CLCannyEdgeKernel.cpp
deleted file mode 100644
index 1fe944c8a2..0000000000
--- a/src/core/CL/kernels/CLCannyEdgeKernel.cpp
+++ /dev/null
@@ -1,310 +0,0 @@
-/*
- * Copyright (c) 2017-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLCannyEdgeKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-using namespace arm_compute;
-
-CLGradientKernel::CLGradientKernel()
-    : _gx(nullptr), _gy(nullptr), _magnitude(nullptr), _phase(nullptr)
-{
-}
-
-void CLGradientKernel::configure(const ICLTensor *gx, const ICLTensor *gy, ICLTensor *magnitude, ICLTensor *phase, int32_t norm_type)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), gx, gy, magnitude, phase, norm_type);
-}
-
-void CLGradientKernel::configure(const CLCompileContext &compile_context, const ICLTensor *gx, const ICLTensor *gy, ICLTensor *magnitude, ICLTensor *phase, int32_t norm_type)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(gx, 1, DataType::S16, DataType::S32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(gy, 1, DataType::S16, DataType::S32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(magnitude, 1, DataType::U16, DataType::U32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(phase, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_MSG(data_size_from_type(gx->info()->data_type()) != data_size_from_type(gy->info()->data_type()),
-                             "Gx and Gy must have the same pixel size");
-    ARM_COMPUTE_ERROR_ON_MSG(data_size_from_type(gx->info()->data_type()) != data_size_from_type(magnitude->info()->data_type()),
-                             "Mag must have the same pixel size as Gx and Gy");
-
-    _gx        = gx;
-    _gy        = gy;
-    _magnitude = magnitude;
-    _phase     = phase;
-
-    // Create build opts
-    std::set<std::string> built_opts;
-    built_opts.emplace("-DDATA_TYPE_IN=" + get_cl_type_from_data_type(gx->info()->data_type()));
-    built_opts.emplace("-DDATA_TYPE_OUT=" + get_cl_type_from_data_type(gx->info()->data_type()));
-
-    // Create kernel
-    const std::string kernel_name = (norm_type == 1) ? std::string("combine_gradients_L1") : std::string("combine_gradients_L2");
-    _kernel                       = create_kernel(compile_context, kernel_name, built_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 4;
-
-    Window win = calculate_max_window(*_gx->info(), Steps(num_elems_processed_per_iteration));
-
-    AccessWindowHorizontal gx_access(_gx->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal gy_access(_gy->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal mag_access(_magnitude->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal phase_access(_phase->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win, gx_access, gy_access, mag_access, phase_access);
-
-    mag_access.set_valid_region(win, _gx->info()->valid_region());
-    phase_access.set_valid_region(win, _gx->info()->valid_region());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(gx->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(gx->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(gx->info()->dimension(1));
-}
-
-void CLGradientKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _gx, slice);
-        add_2D_tensor_argument(idx, _gy, slice);
-        add_2D_tensor_argument(idx, _magnitude, slice);
-        add_2D_tensor_argument(idx, _phase, slice);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
-
-CLEdgeNonMaxSuppressionKernel::CLEdgeNonMaxSuppressionKernel()
-    : _magnitude(nullptr), _phase(nullptr), _output(nullptr)
-{
-}
-
-BorderSize CLEdgeNonMaxSuppressionKernel::border_size() const
-{
-    return BorderSize(1);
-}
-
-void CLEdgeNonMaxSuppressionKernel::configure(const ICLTensor *magnitude, const ICLTensor *phase, ICLTensor *output, int32_t lower_thr, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), magnitude, phase, output, lower_thr, border_undefined);
-}
-
-void CLEdgeNonMaxSuppressionKernel::configure(const CLCompileContext &compile_context, const ICLTensor *magnitude, const ICLTensor *phase, ICLTensor *output, int32_t lower_thr, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(magnitude, 1, DataType::U16, DataType::U32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(phase, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U16, DataType::U32);
-
-    _magnitude = magnitude;
-    _phase     = phase;
-    _output    = output;
-
-    // Create build opts
-    std::set<std::string> built_opts;
-    built_opts.emplace("-DDATA_TYPE_IN=" + get_cl_type_from_data_type(magnitude->info()->data_type()));
-    built_opts.emplace("-DDATA_TYPE_OUT=" + get_cl_type_from_data_type(output->info()->data_type()));
-
-    // Create kernel
-    const std::string kernel_name = std::string("suppress_non_maximum");
-    _kernel                       = create_kernel(compile_context, kernel_name, built_opts);
-
-    // Set minimum threshold argument
-    unsigned int idx = 3 * num_arguments_per_2D_tensor(); //Skip the input and output parameters
-    _kernel.setArg(idx++, lower_thr);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration    = 1;
-    constexpr unsigned int num_elems_read_written_per_iteration = 3;
-
-    Window win = calculate_max_window(*_magnitude->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle mag_access(_magnitude->info(), -border_size().left, -border_size().top,
-                                     num_elems_read_written_per_iteration, num_elems_read_written_per_iteration);
-    AccessWindowHorizontal phase_access(_phase->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal output_access(_output->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win, mag_access, phase_access, output_access);
-
-    output_access.set_valid_region(win, _magnitude->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(output->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(border_undefined);
-}
-
-void CLEdgeNonMaxSuppressionKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _magnitude, slice);
-        add_2D_tensor_argument(idx, _phase, slice);
-        add_2D_tensor_argument(idx, _output, slice);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
-
-CLEdgeTraceKernel::CLEdgeTraceKernel()
-    : _input(nullptr), _output(nullptr), _lower_thr(0), _upper_thr(0), _visited(nullptr), _recorded(nullptr), _l1_stack(nullptr), _l1_stack_counter(nullptr)
-{
-}
-
-void CLEdgeTraceKernel::configure(const ICLTensor *input, ICLTensor *output, int32_t upper_thr, int32_t lower_thr,
-                                  ICLTensor *visited, ICLTensor *recorded, ICLTensor *l1_stack, ICLTensor *l1_stack_counter)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, upper_thr, lower_thr, visited, recorded, l1_stack, l1_stack_counter);
-}
-
-void CLEdgeTraceKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, int32_t upper_thr, int32_t lower_thr,
-                                  ICLTensor *visited, ICLTensor *recorded, ICLTensor *l1_stack, ICLTensor *l1_stack_counter)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U16, DataType::U32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(visited, 1, DataType::U32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(recorded, 1, DataType::U32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(l1_stack, 1, DataType::S32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(l1_stack_counter, 1, DataType::U8);
-
-    _input            = input;
-    _output           = output;
-    _lower_thr        = lower_thr;
-    _upper_thr        = upper_thr;
-    _visited          = visited;
-    _recorded         = recorded;
-    _l1_stack         = l1_stack;
-    _l1_stack_counter = l1_stack_counter;
-
-    // Create build opts
-    std::set<std::string> built_opts;
-    built_opts.emplace("-DDATA_TYPE_IN=" + get_cl_type_from_data_type(input->info()->data_type()));
-    built_opts.emplace("-DDATA_TYPE_OUT=" + get_cl_type_from_data_type(output->info()->data_type()));
-
-    // Create kernel
-    const std::string kernel_name = std::string("hysteresis");
-    _kernel                       = create_kernel(compile_context, kernel_name, built_opts);
-
-    // Set constant kernel args
-    unsigned int width  = _input->info()->dimension(0);
-    unsigned int height = _input->info()->dimension(1);
-    unsigned int idx    = 6 * num_arguments_per_2D_tensor(); //Skip the input and output parameters
-    _kernel.setArg(idx++, static_cast<cl_uint>(_lower_thr));
-    _kernel.setArg(idx++, static_cast<cl_uint>(_upper_thr));
-    _kernel.setArg(idx++, static_cast<cl_uint>(width));
-    _kernel.setArg(idx++, static_cast<cl_uint>(height));
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 1;
-    Window                 win                               = calculate_max_window(*_input->info(), Steps(num_elems_processed_per_iteration));
-
-    AccessWindowHorizontal output_access(_output->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal visited_access(_visited->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal recorded_access(_recorded->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal l1_stack_access(_l1_stack->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal l1_stack_counter_access(_l1_stack_counter->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win,
-                              AccessWindowHorizontal(_input->info(), 0, num_elems_processed_per_iteration),
-                              output_access,
-                              visited_access,
-                              recorded_access,
-                              l1_stack_access,
-                              l1_stack_counter_access);
-
-    output_access.set_valid_region(win, _input->info()->valid_region());
-    visited_access.set_valid_region(win, _input->info()->valid_region());
-    recorded_access.set_valid_region(win, _input->info()->valid_region());
-    l1_stack_access.set_valid_region(win, _input->info()->valid_region());
-    l1_stack_counter_access.set_valid_region(win, _input->info()->valid_region());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += lower_string(string_from_format(output->info()->format()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-}
-
-void CLEdgeTraceKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-        add_2D_tensor_argument(idx, _output, slice);
-        add_2D_tensor_argument(idx, _visited, slice);
-        add_2D_tensor_argument(idx, _recorded, slice);
-        add_2D_tensor_argument(idx, _l1_stack, slice);
-        add_2D_tensor_argument(idx, _l1_stack_counter, slice);
-
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLCannyEdgeKernel.h b/src/core/CL/kernels/CLCannyEdgeKernel.h
deleted file mode 100644
index 7543822d8d..0000000000
--- a/src/core/CL/kernels/CLCannyEdgeKernel.h
+++ /dev/null
@@ -1,185 +0,0 @@
-/*
- * Copyright (c) 2017-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLCANNYEDGEKERNEL_H
-#define ARM_COMPUTE_CLCANNYEDGEKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** OpenCL kernel to perform Gradient computation.
- */
-class CLGradientKernel : public ICLKernel
-{
-public:
-    /** Constructor */
-    CLGradientKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGradientKernel(const CLGradientKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGradientKernel &operator=(const CLGradientKernel &) = delete;
-    /** Initialise the kernel's sources, destinations and border mode.
-     *
-     * @note gx, gy and mag must all be the same size (either 16 or 32).
-     *
-     * @param[in]  gx        Source tensor - Gx component. Data types supported: S16/S32.
-     * @param[in]  gy        Source tensor - Gy component. Data types supported: Same as gx.
-     * @param[out] magnitude Destination tensor - Magnitude. Data types supported: U16/U32. Must match the pixel size of gx, gy.
-     * @param[out] phase     Destination tensor - Quantized phase. Data types supported: U8.
-     * @param[in]  norm_type Normalization type. if 1, L1-Norm otherwise L2-Norm.
-     */
-    void configure(const ICLTensor *gx, const ICLTensor *gy, ICLTensor *magnitude, ICLTensor *phase, int32_t norm_type);
-    /** Initialise the kernel's sources, destinations and border mode.
-     *
-     * @note gx, gy and mag must all be the same size (either 16 or 32).
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  gx              Source tensor - Gx component. Data types supported: S16/S32.
-     * @param[in]  gy              Source tensor - Gy component. Data types supported: Same as gx.
-     * @param[out] magnitude       Destination tensor - Magnitude. Data types supported: U16/U32. Must match the pixel size of gx, gy.
-     * @param[out] phase           Destination tensor - Quantized phase. Data types supported: U8.
-     * @param[in]  norm_type       Normalization type. if 1, L1-Norm otherwise L2-Norm.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *gx, const ICLTensor *gy, ICLTensor *magnitude, ICLTensor *phase, int32_t norm_type);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_gx;        /**< Source tensor - Gx component */
-    const ICLTensor *_gy;        /**< Source tensor - Gy component */
-    ICLTensor       *_magnitude; /**< Destination tensor - Magnitude */
-    ICLTensor       *_phase;     /**< Destination tensor - Quantized phase */
-};
-
-/** OpenCL kernel to perform Non-Maxima suppression for Canny Edge.
- *
- * @note This kernel is meant to be used alongside CannyEdge and performs a non-maxima suppression using magnitude and phase of input
- *       to characterize points as possible edges. The output buffer needs to be cleared before this kernel is executed.
- *
- * @note Hysteresis is computed in @ref CLEdgeTraceKernel
- */
-class CLEdgeNonMaxSuppressionKernel : public ICLKernel
-{
-public:
-    /** Constructor */
-    CLEdgeNonMaxSuppressionKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLEdgeNonMaxSuppressionKernel(const CLEdgeNonMaxSuppressionKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLEdgeNonMaxSuppressionKernel &operator=(const CLEdgeNonMaxSuppressionKernel &) = delete;
-    /** Initialise the kernel's sources, destination and border mode.
-     *
-     * @param[in]  magnitude        Source tensor - Magnitude. Data types supported: U16/U32.
-     * @param[in]  phase            Source tensor - Quantized phase. Data types supported: U8.
-     * @param[out] output           Destination tensor. Data types supported: U16/U32.
-     * @param[in]  lower_thr        Lower threshold.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *magnitude, const ICLTensor *phase, ICLTensor *output, int32_t lower_thr, bool border_undefined);
-    /** Initialise the kernel's sources, destination and border mode.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  magnitude        Source tensor - Magnitude. Data types supported: U16/U32.
-     * @param[in]  phase            Source tensor - Quantized phase. Data types supported: U8.
-     * @param[out] output           Destination tensor. Data types supported: U16/U32.
-     * @param[in]  lower_thr        Lower threshold.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *magnitude, const ICLTensor *phase, ICLTensor *output, int32_t lower_thr, bool border_undefined);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-    BorderSize border_size() const override;
-
-private:
-    const ICLTensor *_magnitude; /**< Source tensor - Magnitude. */
-    const ICLTensor *_phase;     /**< Source tensor - Quantized phase. */
-    ICLTensor       *_output;    /**< Destination tensor. */
-};
-
-/** OpenCL kernel to perform Edge tracing.
- */
-class CLEdgeTraceKernel : public ICLKernel
-{
-public:
-    /** Constructor */
-    CLEdgeTraceKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLEdgeTraceKernel(const CLEdgeTraceKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLEdgeTraceKernel &operator=(const CLEdgeTraceKernel &) = delete;
-    /** Initialise the kernel's source, destination and border mode.
-     *
-     * @param[in]     input            Source tensor. Data types supported: U16/U32.
-     * @param[out]    output           Destination tensor. Data types supported: U8.
-     * @param[in]     upper_thr        Upper threshold used for the hysteresis
-     * @param[in]     lower_thr        Lower threshold used for the hysteresis
-     * @param[in,out] visited          Tensor for keeping the visited pixels. Data types supported: U32.
-     *                                 Expected to be initialized to 0 before each run.
-     * @param[in,out] recorded         Tensor for keeping the recorded pixels. Data types supported: U32
-     *                                 Expected to be initialized to 0 before each run.
-     * @param[in,out] l1_stack         Tensor with the L1 stack for each pixel. Data types supported: S32.
-     *                                 Expected to be initialized to 0 before each run.
-     * @param[in,out] l1_stack_counter Tensor for counting the elements in the L1 stack of each pixel. Data types supported: U8.
-     *                                              Expected to be initialized to 0 before each run.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, int32_t upper_thr, int32_t lower_thr,
-                   ICLTensor *visited, ICLTensor *recorded, ICLTensor *l1_stack, ICLTensor *l1_stack_counter);
-    /** Initialise the kernel's source, destination and border mode.
-     *
-     * @param[in]     compile_context  The compile context to be used.
-     * @param[in]     input            Source tensor. Data types supported: U16/U32.
-     * @param[out]    output           Destination tensor. Data types supported: U8.
-     * @param[in]     upper_thr        Upper threshold used for the hysteresis
-     * @param[in]     lower_thr        Lower threshold used for the hysteresis
-     * @param[in,out] visited          Tensor for keeping the visited pixels. Data types supported: U32.
-     *                                 Expected to be initialized to 0 before each run.
-     * @param[in,out] recorded         Tensor for keeping the recorded pixels. Data types supported: U32
-     *                                 Expected to be initialized to 0 before each run.
-     * @param[in,out] l1_stack         Tensor with the L1 stack for each pixel. Data types supported: S32.
-     *                                 Expected to be initialized to 0 before each run.
-     * @param[in,out] l1_stack_counter Tensor for counting the elements in the L1 stack of each pixel. Data types supported: U8.
-     *                                              Expected to be initialized to 0 before each run.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, int32_t upper_thr, int32_t lower_thr,
-                   ICLTensor *visited, ICLTensor *recorded, ICLTensor *l1_stack, ICLTensor *l1_stack_counter);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_input;            /**< Source tensor. */
-    ICLTensor       *_output;           /**< Destination tensor. */
-    int32_t          _lower_thr;        /**< Lower threshold used for the hysteresis. */
-    int32_t          _upper_thr;        /**< Upper threshold used for the hysteresis. */
-    ICLTensor       *_visited;          /**< Marks visited elements */
-    ICLTensor       *_recorded;         /**< Marks recorded elements */
-    ICLTensor       *_l1_stack;         /**< L1 hysteris stack */
-    ICLTensor       *_l1_stack_counter; /**< L1 hysteris stack counter */
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLCANNYEDGEKERNEL_H */
diff --git a/src/core/CL/kernels/CLChannelCombineKernel.cpp b/src/core/CL/kernels/CLChannelCombineKernel.cpp
deleted file mode 100644
index 52ba9dd065..0000000000
--- a/src/core/CL/kernels/CLChannelCombineKernel.cpp
+++ /dev/null
@@ -1,296 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLChannelCombineKernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLMultiImage.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/MultiImageInfo.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-
-#include <set>
-#include <string>
-
-namespace arm_compute
-{
-namespace
-{
-constexpr unsigned int num_elems_processed_per_iteration = 16;
-} // namespace
-
-CLChannelCombineKernel::CLChannelCombineKernel()
-    : _planes{ { nullptr } }, _output(nullptr), _output_multi(nullptr), _x_subsampling{ { 1, 1, 1 } }, _y_subsampling{ { 1, 1, 1 } }
-{
-}
-
-void CLChannelCombineKernel::configure(const ICLTensor *plane0, const ICLTensor *plane1, const ICLTensor *plane2, const ICLTensor *plane3, ICLTensor *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), plane0, plane1, plane2, plane3, output);
-}
-
-void CLChannelCombineKernel::configure(const CLCompileContext &compile_context, const ICLTensor *plane0, const ICLTensor *plane1, const ICLTensor *plane2, const ICLTensor *plane3, ICLTensor *output)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(plane0, plane1, plane2, output);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(plane0);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(plane1);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(plane2);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(output);
-
-    ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(plane0, Format::U8);
-    ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(plane1, Format::U8);
-    ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(plane2, Format::U8);
-    ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(output, Format::RGB888, Format::RGBA8888, Format::YUYV422, Format::UYVY422);
-
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(plane0, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(plane1, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(plane2, 1, DataType::U8);
-
-    const Format output_format = output->info()->format();
-
-    // Check if horizontal dimension of Y plane is even and validate horizontal sub-sampling dimensions for U and V planes
-    if(Format::YUYV422 == output_format || Format::UYVY422 == output_format)
-    {
-        // Validate Y plane of input and output
-        ARM_COMPUTE_ERROR_ON_TENSORS_NOT_EVEN(output_format, plane0, output);
-
-        // Validate U and V plane of the input
-        ARM_COMPUTE_ERROR_ON_TENSORS_NOT_SUBSAMPLED(output_format, plane0->info()->tensor_shape(), plane1, plane2);
-    }
-
-    _planes[0] = plane0;
-    _planes[1] = plane1;
-    _planes[2] = plane2;
-    _planes[3] = nullptr;
-
-    // Validate the last input tensor only for RGBA format
-    if(Format::RGBA8888 == output_format)
-    {
-        ARM_COMPUTE_ERROR_ON_NULLPTR(plane3);
-        ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(plane3);
-
-        ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(plane3, Format::U8);
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(plane3, 1, DataType::U8);
-
-        _planes[3] = plane3;
-    }
-
-    _output       = output;
-    _output_multi = nullptr;
-
-    // Half the processed elements for U and V channels due to horizontal sub-sampling of 2
-    if(Format::YUYV422 == output_format || Format::UYVY422 == output_format)
-    {
-        _x_subsampling[1] = 2;
-        _x_subsampling[2] = 2;
-    }
-
-    // Create kernel
-    std::string kernel_name = "channel_combine_" + string_from_format(output_format);
-    _kernel                 = create_kernel(compile_context, kernel_name);
-
-    // Configure window
-    Window win = calculate_max_window(*output->info(), Steps(num_elems_processed_per_iteration));
-
-    AccessWindowHorizontal plane0_access(plane0->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowRectangle  plane1_access(plane1->info(), 0, 0, num_elems_processed_per_iteration, 1, 1.f / _x_subsampling[1], 1.f / _y_subsampling[1]);
-    AccessWindowRectangle  plane2_access(plane2->info(), 0, 0, num_elems_processed_per_iteration, 1, 1.f / _x_subsampling[2], 1.f / _y_subsampling[2]);
-    AccessWindowHorizontal plane3_access(plane3 == nullptr ? nullptr : plane3->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win, plane0_access, plane1_access, plane2_access, plane3_access, output_access);
-
-    ValidRegion valid_region = intersect_valid_regions(plane0->info()->valid_region(),
-                                                       plane1->info()->valid_region(),
-                                                       plane2->info()->valid_region());
-    if(plane3 != nullptr)
-    {
-        valid_region = intersect_valid_regions(plane3->info()->valid_region(), valid_region);
-    }
-    output_access.set_valid_region(win, ValidRegion(valid_region.anchor, output->info()->tensor_shape()));
-
-    ICLKernel::configure_internal(win);
-}
-
-void CLChannelCombineKernel::configure(const ICLImage *plane0, const ICLImage *plane1, const ICLImage *plane2, ICLMultiImage *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), plane0, plane1, plane2, output);
-}
-
-void CLChannelCombineKernel::configure(const CLCompileContext &compile_context, const ICLImage *plane0, const ICLImage *plane1, const ICLImage *plane2, ICLMultiImage *output)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(plane0, plane1, plane2, output);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(plane0);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(plane1);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(plane2);
-
-    ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(plane0, Format::U8);
-    ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(plane1, Format::U8);
-    ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(plane2, Format::U8);
-    ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(output, Format::NV12, Format::NV21, Format::IYUV, Format::YUV444);
-
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(plane0, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(plane1, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(plane2, 1, DataType::U8);
-
-    const Format output_format = output->info()->format();
-
-    // Validate shape of Y plane to be even and shape of sub-sampling dimensions for U and V planes
-    // Perform validation only for formats which require sub-sampling.
-    if(Format::YUV444 != output_format)
-    {
-        // Validate Y plane of input and output
-        ARM_COMPUTE_ERROR_ON_TENSORS_NOT_EVEN(output_format, plane0, output->plane(0));
-
-        // Validate U and V plane of the input
-        ARM_COMPUTE_ERROR_ON_TENSORS_NOT_SUBSAMPLED(output_format, plane0->info()->tensor_shape(), plane1, plane2);
-
-        // Validate second plane U (NV12 and NV21 have a UV88 combined plane while IYUV has only the U plane)
-        // MultiImage generates the correct tensor shape but also check in case the tensor shape of planes was changed to a wrong size
-        ARM_COMPUTE_ERROR_ON_TENSORS_NOT_SUBSAMPLED(output_format, plane0->info()->tensor_shape(), output->plane(1));
-
-        // Validate the last plane V of format IYUV
-        if(Format::IYUV == output_format)
-        {
-            // Validate Y plane of the output
-            ARM_COMPUTE_ERROR_ON_TENSORS_NOT_SUBSAMPLED(output_format, plane0->info()->tensor_shape(), output->plane(2));
-        }
-    }
-
-    // Set input tensors
-    _planes[0] = plane0;
-    _planes[1] = plane1;
-    _planes[2] = plane2;
-    _planes[3] = nullptr;
-
-    // Set output tensor
-    _output       = nullptr;
-    _output_multi = output;
-
-    bool has_two_planars = false;
-
-    // Set sub-sampling parameters for each plane
-    std::string           kernel_name;
-    std::set<std::string> build_opts;
-
-    if(Format::NV12 == output_format || Format::NV21 == output_format)
-    {
-        _x_subsampling = { { 1, 2, 2 } };
-        _y_subsampling = { { 1, 2, 2 } };
-        kernel_name    = "channel_combine_NV";
-        build_opts.emplace(Format::NV12 == output_format ? "-DNV12" : "-DNV21");
-        has_two_planars = true;
-    }
-    else
-    {
-        if(Format::IYUV == output_format)
-        {
-            _x_subsampling = { { 1, 2, 2 } };
-            _y_subsampling = { { 1, 2, 2 } };
-        }
-
-        kernel_name = "copy_planes_3p";
-        build_opts.emplace(Format::IYUV == output_format ? "-DIYUV" : "-DYUV444");
-    }
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, kernel_name, build_opts);
-
-    // Configure window
-    Window win = calculate_max_window(*plane0->info(), Steps(num_elems_processed_per_iteration));
-
-    AccessWindowRectangle input_plane0_access(plane0->info(), 0, 0, num_elems_processed_per_iteration, 1.f);
-    AccessWindowRectangle input_plane1_access(plane1->info(), 0, 0, num_elems_processed_per_iteration, 1.f, 1.f / _x_subsampling[1], 1.f / _y_subsampling[1]);
-    AccessWindowRectangle input_plane2_access(plane2->info(), 0, 0, num_elems_processed_per_iteration, 1.f, 1.f / _x_subsampling[2], 1.f / _y_subsampling[2]);
-    AccessWindowRectangle output_plane0_access(output->plane(0)->info(), 0, 0, num_elems_processed_per_iteration, 1.f, 1.f, 1.f / _y_subsampling[1]);
-    AccessWindowRectangle output_plane1_access(output->plane(1)->info(), 0, 0, num_elems_processed_per_iteration, 1.f, 1.f / _x_subsampling[1], 1.f / _y_subsampling[1]);
-    AccessWindowRectangle output_plane2_access(has_two_planars ? nullptr : output->plane(2)->info(), 0, 0, num_elems_processed_per_iteration, 1.f, 1.f / _x_subsampling[2], 1.f / _y_subsampling[2]);
-
-    update_window_and_padding(win,
-                              input_plane0_access, input_plane1_access, input_plane2_access,
-                              output_plane0_access, output_plane1_access, output_plane2_access);
-
-    ValidRegion plane0_valid_region  = plane0->info()->valid_region();
-    ValidRegion output_plane1_region = has_two_planars ? intersect_valid_regions(plane1->info()->valid_region(), plane2->info()->valid_region()) : plane2->info()->valid_region();
-    output_plane0_access.set_valid_region(win, ValidRegion(plane0_valid_region.anchor, output->plane(0)->info()->tensor_shape()));
-    output_plane1_access.set_valid_region(win, ValidRegion(output_plane1_region.anchor, output->plane(1)->info()->tensor_shape()));
-    output_plane2_access.set_valid_region(win, ValidRegion(plane2->info()->valid_region().anchor, output->plane(2)->info()->tensor_shape()));
-
-    ICLKernel::configure_internal(win);
-}
-
-void CLChannelCombineKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    slice.set_dimension_step(Window::DimY, 1);
-
-    do
-    {
-        // Subsampling in plane 1
-        Window win_sub_plane1(slice);
-        win_sub_plane1.set(Window::DimX, Window::Dimension(win_sub_plane1.x().start() / _x_subsampling[1], win_sub_plane1.x().end() / _x_subsampling[1], win_sub_plane1.x().step() / _x_subsampling[1]));
-        win_sub_plane1.set(Window::DimY, Window::Dimension(win_sub_plane1.y().start() / _y_subsampling[1], win_sub_plane1.y().end() / _y_subsampling[1], 1));
-
-        // Subsampling in plane 2
-        Window win_sub_plane2(slice);
-        win_sub_plane2.set(Window::DimX, Window::Dimension(win_sub_plane2.x().start() / _x_subsampling[2], win_sub_plane2.x().end() / _x_subsampling[2], win_sub_plane2.x().step() / _x_subsampling[2]));
-        win_sub_plane2.set(Window::DimY, Window::Dimension(win_sub_plane2.y().start() / _y_subsampling[2], win_sub_plane2.y().end() / _y_subsampling[2], 1));
-
-        unsigned int idx = 0;
-
-        // Set inputs
-        add_2D_tensor_argument(idx, _planes[0], slice);
-        add_2D_tensor_argument(idx, _planes[1], win_sub_plane1);
-        add_2D_tensor_argument(idx, _planes[2], win_sub_plane2);
-        add_2D_tensor_argument_if((nullptr != _planes[3]), idx, _planes[3], slice);
-
-        // Set outputs
-        if(nullptr != _output) // Single planar output
-        {
-            add_2D_tensor_argument(idx, _output, slice);
-        }
-        else // Multi-planar output
-        {
-            // Reduce slice in case of subsampling to avoid out-of bounds access
-            slice.set(Window::DimY, Window::Dimension(slice.y().start() / _y_subsampling[1], slice.y().end() / _y_subsampling[1], 1));
-
-            add_2D_tensor_argument(idx, _output_multi->cl_plane(0), slice);
-            add_2D_tensor_argument(idx, _output_multi->cl_plane(1), win_sub_plane1);
-            add_2D_tensor_argument_if((3 == num_planes_from_format(_output_multi->info()->format())), idx, _output_multi->cl_plane(2), win_sub_plane2);
-
-            _kernel.setArg(idx++, slice.y().end());
-        }
-
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLChannelCombineKernel.h b/src/core/CL/kernels/CLChannelCombineKernel.h
deleted file mode 100644
index f19995aa8e..0000000000
--- a/src/core/CL/kernels/CLChannelCombineKernel.h
+++ /dev/null
@@ -1,102 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLCHANNELCOMBINEKERNEL_H
-#define ARM_COMPUTE_CLCHANNELCOMBINEKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-#include <array>
-#include <cstdint>
-
-namespace arm_compute
-{
-class ICLMultiImage;
-class ICLTensor;
-using ICLImage = ICLTensor;
-
-/** Interface for the channel combine kernel */
-class CLChannelCombineKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLChannelCombineKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLChannelCombineKernel(const CLChannelCombineKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLChannelCombineKernel &operator=(const CLChannelCombineKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLChannelCombineKernel(CLChannelCombineKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLChannelCombineKernel &operator=(CLChannelCombineKernel &&) = default;
-    /** Default destructor */
-    ~CLChannelCombineKernel() = default;
-    /** Configure function's inputs and outputs.
-     *
-     * @param[in]  plane0 The 2D plane that forms channel 0. Must be of U8 format.
-     * @param[in]  plane1 The 2D plane that forms channel 1. Must be of U8 format.
-     * @param[in]  plane2 The 2D plane that forms channel 2. Must be of U8 format.
-     * @param[in]  plane3 The 2D plane that forms channel 3. Must be of U8 format.
-     * @param[out] output The single planar output tensor. Supported formats: RGB888/RGBA8888/YUYV422/UYVY422.
-     */
-    void configure(const ICLTensor *plane0, const ICLTensor *plane1, const ICLTensor *plane2, const ICLTensor *plane3, ICLTensor *output);
-    /** Configure function's inputs and outputs.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  plane0          The 2D plane that forms channel 0. Must be of U8 format.
-     * @param[in]  plane1          The 2D plane that forms channel 1. Must be of U8 format.
-     * @param[in]  plane2          The 2D plane that forms channel 2. Must be of U8 format.
-     * @param[in]  plane3          The 2D plane that forms channel 3. Must be of U8 format.
-     * @param[out] output          The single planar output tensor.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *plane0, const ICLTensor *plane1, const ICLTensor *plane2, const ICLTensor *plane3, ICLTensor *output);
-    /** Configure function's inputs and outputs.
-     *
-     * @param[in]  plane0 The 2D plane that forms channel 0. Must be of U8 format.
-     * @param[in]  plane1 The 2D plane that forms channel 1. Must be of U8 format.
-     * @param[in]  plane2 The 2D plane that forms channel 2. Must be of U8 format.
-     * @param[out] output The multi planar output tensor. Supported formats: RGB888/RGBA8888/YUYV422/UYVY422.
-     */
-    void configure(const ICLImage *plane0, const ICLImage *plane1, const ICLImage *plane2, ICLMultiImage *output);
-    /** Configure function's inputs and outputs.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  plane0          The 2D plane that forms channel 0. Must be of U8 format.
-     * @param[in]  plane1          The 2D plane that forms channel 1. Must be of U8 format.
-     * @param[in]  plane2          The 2D plane that forms channel 2. Must be of U8 format.
-     * @param[out] output          The multi planar output tensor. Supported formats: RGB888/RGBA8888/YUYV422/UYVY422.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLImage *plane0, const ICLImage *plane1, const ICLImage *plane2, ICLMultiImage *output);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    std::array<const ICLTensor *, 4> _planes;
-    ICLTensor     *_output;
-    ICLMultiImage *_output_multi;
-    std::array<uint32_t, 3> _x_subsampling;
-    std::array<uint32_t, 3> _y_subsampling;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLCHANNELCOMBINEKERNEL_H */
diff --git a/src/core/CL/kernels/CLChannelExtractKernel.cpp b/src/core/CL/kernels/CLChannelExtractKernel.cpp
deleted file mode 100644
index cbf504b98b..0000000000
--- a/src/core/CL/kernels/CLChannelExtractKernel.cpp
+++ /dev/null
@@ -1,196 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLChannelExtractKernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLMultiImage.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Coordinates.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/MultiImageInfo.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/AutoConfiguration.h"
-#include "src/core/helpers/WindowHelpers.h"
-
-#include <set>
-#include <string>
-
-using namespace arm_compute;
-
-CLChannelExtractKernel::CLChannelExtractKernel()
-    : _input(nullptr), _output(nullptr), _num_elems_processed_per_iteration(8), _subsampling(1)
-{
-}
-
-void CLChannelExtractKernel::configure(const ICLTensor *input, Channel channel, ICLTensor *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, channel, output);
-}
-
-void CLChannelExtractKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, Channel channel, ICLTensor *output)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_ERROR_ON(input == output);
-
-    set_format_if_unknown(*output->info(), Format::U8);
-
-    // Check if input tensor has a valid format
-    ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(input, Format::RGB888, Format::RGBA8888, Format::YUYV422, Format::UYVY422);
-    ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(output, Format::U8);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(output);
-
-    // Check if channel is valid for given format
-    const Format format = input->info()->format();
-    ARM_COMPUTE_ERROR_ON_CHANNEL_NOT_IN_KNOWN_FORMAT(format, channel);
-
-    // Half the processed elements for U,V channels due to sub-sampling of 2
-    _subsampling = 1;
-
-    if(format == Format::YUYV422 || format == Format::UYVY422)
-    {
-        // Check if the width of the tensor shape is even for formats with subsampled channels (UYVY422 and YUYV422)
-        ARM_COMPUTE_ERROR_ON_TENSORS_NOT_EVEN(format, input);
-
-        if(channel != Channel::Y)
-        {
-            _subsampling = 2;
-        }
-    }
-
-    // Calculate output tensor shape using subsampling
-    TensorShape output_shape = calculate_subsampled_shape(input->info()->tensor_shape(), format, channel);
-    set_shape_if_empty(*output->info(), output_shape);
-
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(output->info()->tensor_shape(), output_shape);
-
-    _input  = input;
-    _output = output;
-
-    // Create kernel
-    std::string           kernel_name = "channel_extract_" + string_from_format(format);
-    std::set<std::string> build_opts  = { ("-DCHANNEL_" + string_from_channel(channel)) };
-    _kernel                           = create_kernel(compile_context, kernel_name, build_opts);
-
-    // Configure window
-    Window                 win = calculate_max_window(*input->info(), Steps(_num_elems_processed_per_iteration));
-    AccessWindowHorizontal input_access(input->info(), 0, _num_elems_processed_per_iteration);
-    AccessWindowRectangle  output_access(output->info(), 0, 0, _num_elems_processed_per_iteration, 1, 1.f / _subsampling, 1.f / _subsampling);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    ValidRegion input_valid_region = input->info()->valid_region();
-    output_access.set_valid_region(win, ValidRegion(input_valid_region.anchor, output->info()->tensor_shape()));
-
-    ICLKernel::configure_internal(win);
-}
-
-void CLChannelExtractKernel::configure(const ICLMultiImage *input, Channel channel, ICLImage *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, channel, output);
-}
-
-void CLChannelExtractKernel::configure(const CLCompileContext &compile_context, const ICLMultiImage *input, Channel channel, ICLImage *output)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(output);
-
-    set_format_if_unknown(*output->info(), Format::U8);
-
-    // Check if channel is valid for given format
-    const Format format = input->info()->format();
-    ARM_COMPUTE_ERROR_ON_CHANNEL_NOT_IN_KNOWN_FORMAT(format, channel);
-
-    // Get input plane from the given channel
-    const ICLImage *input_plane = input->cl_plane(plane_idx_from_channel(format, channel));
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input_plane);
-
-    if(Channel::Y == channel && format != Format::YUV444)
-    {
-        // Check if the width of the tensor shape is even for formats with subsampled channels (UYVY422 and YUYV422)
-        ARM_COMPUTE_ERROR_ON_TENSORS_NOT_EVEN(format, input_plane);
-    }
-
-    // Calculate 2x2 subsampled tensor shape
-    TensorShape output_shape = calculate_subsampled_shape(input->cl_plane(0)->info()->tensor_shape(), format, channel);
-    set_shape_if_empty(*output->info(), output_shape);
-
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(output_shape, output->info()->tensor_shape());
-
-    // Check if input tensor has a valid format
-    ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(input, Format::NV12, Format::NV21, Format::IYUV, Format::YUV444);
-    ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(output, Format::U8);
-
-    _output      = output;
-    _input       = input_plane;
-    _subsampling = 1;
-
-    // Create kernel
-    std::string           kernel_name;
-    std::set<std::string> build_opts;
-    if(Channel::Y == channel || Format::IYUV == format || Format::YUV444 == format)
-    {
-        kernel_name = "copy_plane";
-    }
-    else
-    {
-        kernel_name = "channel_extract_" + string_from_format(format);
-        build_opts.insert(("-DCHANNEL_" + string_from_channel(channel)));
-    }
-    _kernel = create_kernel(compile_context, kernel_name, build_opts);
-
-    // Configure window
-    Window                 win = calculate_max_window(*input_plane->info(), Steps(_num_elems_processed_per_iteration));
-    AccessWindowHorizontal input_access(input_plane->info(), 0, _num_elems_processed_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, _num_elems_processed_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input_plane->info()->valid_region());
-
-    ICLKernel::configure_internal(win);
-}
-
-void CLChannelExtractKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-
-    do
-    {
-        Window win_sub(slice);
-        win_sub.set(Window::DimX, Window::Dimension(win_sub.x().start() / _subsampling, win_sub.x().end() / _subsampling, win_sub.x().step() / _subsampling));
-        win_sub.set(Window::DimY, Window::Dimension(win_sub.y().start() / _subsampling, win_sub.y().end() / _subsampling, 1));
-
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-        add_2D_tensor_argument(idx, _output, win_sub);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLChannelExtractKernel.h b/src/core/CL/kernels/CLChannelExtractKernel.h
deleted file mode 100644
index 37abde548c..0000000000
--- a/src/core/CL/kernels/CLChannelExtractKernel.h
+++ /dev/null
@@ -1,95 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLCHANNELEXTRACTKERNEL_H
-#define ARM_COMPUTE_CLCHANNELEXTRACTKERNEL_H
-
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLKernel.h"
-
-#include <cstdint>
-
-namespace arm_compute
-{
-class ICLMultiImage;
-class ICLTensor;
-using ICLImage = ICLTensor;
-
-/** Interface for the channel extract kernel */
-class CLChannelExtractKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLChannelExtractKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLChannelExtractKernel(const CLChannelExtractKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLChannelExtractKernel &operator=(const CLChannelExtractKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLChannelExtractKernel(CLChannelExtractKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLChannelExtractKernel &operator=(CLChannelExtractKernel &&) = default;
-    /** Default destructor */
-    ~CLChannelExtractKernel() = default;
-    /** Set the input and output of the kernel
-     *
-     * @param[in]  input   Source tensor. Formats supported: RGB888/RGBA8888/YUYV422/UYVY422
-     * @param[in]  channel Channel to extract.
-     * @param[out] output  Destination tensor. Must be of U8 format.
-     */
-    void configure(const ICLTensor *input, Channel channel, ICLTensor *output);
-    /** Set the input and output of the kernel
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Source tensor. Formats supported: RGB888/RGBA8888/YUYV422/UYVY422
-     * @param[in]  channel         Channel to extract.
-     * @param[out] output          Destination tensor. Must be of U8 format.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, Channel channel, ICLTensor *output);
-    /** Set the input and output of the kernel
-     *
-     * @param[in]  input   Multi-planar source image. Formats supported: NV12/NV21/IYUV/YUV444
-     * @param[in]  channel Channel to extract.
-     * @param[out] output  Single-planar 2D destination image. Must be of U8 format.
-     */
-    void configure(const ICLMultiImage *input, Channel channel, ICLImage *output);
-    /** Set the input and output of the kernel
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Multi-planar source image. Formats supported: NV12/NV21/IYUV/YUV444
-     * @param[in]  channel         Channel to extract.
-     * @param[out] output          Single-planar 2D destination image. Must be of U8 format.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLMultiImage *input, Channel channel, ICLImage *output);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_input;
-    ICLTensor       *_output;
-    uint32_t         _num_elems_processed_per_iteration;
-    uint32_t         _subsampling;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLCHANNELEXTRACTKERNEL_H */
diff --git a/src/core/CL/kernels/CLColorConvertKernel.cpp b/src/core/CL/kernels/CLColorConvertKernel.cpp
deleted file mode 100644
index 6c61fec997..0000000000
--- a/src/core/CL/kernels/CLColorConvertKernel.cpp
+++ /dev/null
@@ -1,558 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLColorConvertKernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLMultiImage.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/MultiImageInfo.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <sstream>
-
-using namespace arm_compute;
-
-CLColorConvertKernel::CLColorConvertKernel()
-    : _input(nullptr), _output(nullptr), _multi_input(nullptr), _multi_output(nullptr)
-{
-}
-
-void CLColorConvertKernel::configure(const ICLTensor *input, ICLTensor *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output);
-}
-
-void CLColorConvertKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output)
-{
-    ARM_COMPUTE_ERROR_ON(input == nullptr);
-    ARM_COMPUTE_ERROR_ON(output == nullptr);
-
-    unsigned int num_elems_processed_per_iteration = 0;
-    switch(input->info()->format())
-    {
-        case Format::RGBA8888:
-        {
-            switch(output->info()->format())
-            {
-                case Format::RGB888:
-                    num_elems_processed_per_iteration = 16;
-                    break;
-                default:
-                    ARM_COMPUTE_ERROR("Not supported");
-                    break;
-            }
-            break;
-        }
-        case Format::UYVY422:
-        case Format::YUYV422:
-        {
-            switch(output->info()->format())
-            {
-                case Format::RGB888:
-                case Format::RGBA8888:
-                    num_elems_processed_per_iteration = 8;
-                    break;
-                default:
-                    ARM_COMPUTE_ERROR("Not supported");
-                    break;
-            }
-            break;
-        }
-        case Format::RGB888:
-        {
-            switch(output->info()->format())
-            {
-                case Format::RGBA8888:
-                case Format::U8:
-                    num_elems_processed_per_iteration = 16;
-                    break;
-                default:
-                    ARM_COMPUTE_ERROR("Not supported");
-                    break;
-            }
-            break;
-        }
-        default:
-            break;
-    }
-    ARM_COMPUTE_ERROR_ON_MSG_VAR(num_elems_processed_per_iteration == 0, "Conversion from %s to %s not supported",
-                                 string_from_format(input->info()->format()).c_str(),
-                                 string_from_format(output->info()->format()).c_str());
-
-    std::stringstream kernel_name;
-
-    kernel_name << string_from_format(input->info()->format());
-    kernel_name << "_to_";
-    kernel_name << string_from_format(output->info()->format());
-    kernel_name << "_bt709";
-
-    _input  = input;
-    _output = output;
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, kernel_name.str());
-
-    // Configure kernel window
-    Window                 win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration));
-    AccessWindowHorizontal input_access(input->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name.str();
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-}
-
-void CLColorConvertKernel::configure(const ICLMultiImage *input, ICLImage *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output);
-}
-
-void CLColorConvertKernel::configure(const CLCompileContext &compile_context, const ICLMultiImage *input, ICLImage *output)
-{
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(output);
-    ARM_COMPUTE_ERROR_ON(output == nullptr);
-
-    unsigned int num_elems_processed_per_iteration = 0;
-
-    switch(input->info()->format())
-    {
-        case Format::NV12:
-        case Format::NV21:
-        case Format::IYUV:
-        {
-            switch(output->info()->format())
-            {
-                case Format::RGB888:
-                case Format::RGBA8888:
-                    num_elems_processed_per_iteration = 4;
-                    break;
-                default:
-                    ARM_COMPUTE_ERROR("Not supported");
-                    break;
-            }
-            break;
-        }
-        default:
-            break;
-    }
-    ARM_COMPUTE_ERROR_ON_MSG_VAR(num_elems_processed_per_iteration == 0, "Conversion from %s to %s not supported",
-                                 string_from_format(input->info()->format()).c_str(),
-                                 string_from_format(output->info()->format()).c_str());
-
-    std::stringstream kernel_name;
-
-    kernel_name << string_from_format(input->info()->format());
-    kernel_name << "_to_";
-    kernel_name << string_from_format(output->info()->format());
-    kernel_name << "_bt709";
-
-    _multi_input = input;
-    _output      = output;
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, kernel_name.str());
-
-    // Configure kernel window
-    const bool  has_two_planes = (input->info()->format() == Format::NV12) || (input->info()->format() == Format::NV21);
-    const float sub_sampling   = (has_two_planes || (input->info()->format() == Format::IYUV)) ? 0.5f : 1;
-
-    Window win = calculate_max_window(*output->info(), Steps(num_elems_processed_per_iteration));
-    win.set_dimension_step(Window::DimY, 2);
-
-    AccessWindowHorizontal plane0_access(input->plane(0)->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowRectangle  plane1_access(input->plane(1)->info(), 0, 0, num_elems_processed_per_iteration, 1,
-                                         sub_sampling, sub_sampling);
-    AccessWindowRectangle plane2_access(has_two_planes ? nullptr : input->plane(2)->info(), 0, 0, num_elems_processed_per_iteration, 1,
-                                        sub_sampling, sub_sampling);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win,
-                              plane0_access, plane1_access, plane2_access,
-                              output_access);
-
-    ValidRegion intersect_region = intersect_valid_regions(input->plane(0)->info()->valid_region(), input->plane(1)->info()->valid_region(),
-                                                           input->plane(2)->info()->valid_region());
-    output_access.set_valid_region(win, ValidRegion(intersect_region.anchor, output->info()->tensor_shape()));
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name.str();
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->plane(0)->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->plane(0)->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->plane(0)->info()->dimension(1));
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->plane(1)->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->plane(1)->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->plane(1)->info()->dimension(1));
-}
-
-void CLColorConvertKernel::configure(const ICLImage *input, ICLMultiImage *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output);
-}
-
-void CLColorConvertKernel::configure(const CLCompileContext &compile_context, const ICLImage *input, ICLMultiImage *output)
-{
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(input);
-    ARM_COMPUTE_ERROR_ON(output == nullptr);
-
-    unsigned int num_elems_processed_per_iteration = 0;
-    unsigned int num_elems_read_per_iteration_x    = 0;
-
-    bool  has_two_planes = (output->info()->format() == Format::NV12) || (output->info()->format() == Format::NV21);
-    float sub_sampling   = (has_two_planes || (output->info()->format() == Format::IYUV)) ? 0.5f : 1;
-
-    switch(input->info()->format())
-    {
-        case Format::RGB888:
-        case Format::RGBA8888:
-        {
-            switch(output->info()->format())
-            {
-                case Format::NV12:
-                case Format::IYUV:
-                    num_elems_processed_per_iteration = 2;
-                    num_elems_read_per_iteration_x    = 8;
-                    break;
-                case Format::YUV444:
-                    num_elems_processed_per_iteration = 4;
-                    num_elems_read_per_iteration_x    = 16;
-                    break;
-                default:
-                    ARM_COMPUTE_ERROR("Not supported");
-                    break;
-            }
-            break;
-        }
-        case Format::UYVY422:
-        case Format::YUYV422:
-        {
-            switch(output->info()->format())
-            {
-                case Format::NV12:
-                case Format::IYUV:
-                    num_elems_processed_per_iteration = 8;
-                    num_elems_read_per_iteration_x    = 8;
-                    break;
-                default:
-                    ARM_COMPUTE_ERROR("Not supported");
-                    break;
-            }
-            break;
-        }
-        default:
-            break;
-    }
-
-    ARM_COMPUTE_ERROR_ON_MSG_VAR(num_elems_processed_per_iteration == 0, "Conversion from %s to %s not supported",
-                                 string_from_format(input->info()->format()).c_str(),
-                                 string_from_format(output->info()->format()).c_str());
-
-    std::stringstream kernel_name;
-
-    kernel_name << string_from_format(input->info()->format());
-    kernel_name << "_to_";
-    kernel_name << string_from_format(output->info()->format());
-    kernel_name << "_bt709";
-    _input        = input;
-    _multi_output = output;
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, kernel_name.str());
-
-    // Configure kernel window
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration));
-    if((input->info()->format() != Format::RGB888 || output->info()->format() != Format::YUV444) && (input->info()->format() != Format::RGBA8888 || output->info()->format() != Format::YUV444))
-    {
-        win.set_dimension_step(Window::DimY, 2);
-    }
-
-    AccessWindowHorizontal output_plane0_access(output->plane(0)->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowRectangle  output_plane1_access(output->plane(1)->info(), 0, 0, num_elems_processed_per_iteration, 1, sub_sampling, sub_sampling);
-    AccessWindowRectangle  output_plane2_access(has_two_planes ? nullptr : output->plane(2)->info(), 0, 0,
-                                                num_elems_processed_per_iteration, 1, sub_sampling, sub_sampling);
-
-    AccessWindowHorizontal input_access(input->info(), 0, num_elems_read_per_iteration_x);
-
-    update_window_and_padding(win,
-                              input_access,
-                              output_plane0_access,
-                              output_plane1_access,
-                              output_plane2_access);
-
-    ValidRegion input_region = input->info()->valid_region();
-
-    output_plane0_access.set_valid_region(win, ValidRegion(input_region.anchor, output->plane(0)->info()->tensor_shape()));
-    output_plane1_access.set_valid_region(win, ValidRegion(input_region.anchor, output->plane(1)->info()->tensor_shape()));
-    output_plane2_access.set_valid_region(win, ValidRegion(input_region.anchor, output->plane(2)->info()->tensor_shape()));
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name.str();
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-}
-
-void CLColorConvertKernel::configure(const ICLMultiImage *input, ICLMultiImage *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output);
-}
-
-void CLColorConvertKernel::configure(const CLCompileContext &compile_context, const ICLMultiImage *input, ICLMultiImage *output)
-{
-    unsigned int num_elems_processed_per_iteration = 0;
-    switch(input->info()->format())
-    {
-        case Format::NV12:
-        case Format::NV21:
-        {
-            switch(output->info()->format())
-            {
-                case Format::IYUV:
-                case Format::YUV444:
-                    num_elems_processed_per_iteration = 16;
-                    break;
-                default:
-                    ARM_COMPUTE_ERROR("Not supported");
-                    break;
-            }
-            break;
-        }
-        case Format::IYUV:
-        {
-            switch(output->info()->format())
-            {
-                case Format::YUV444:
-                case Format::NV12:
-                    num_elems_processed_per_iteration = 16;
-                    break;
-                default:
-                    ARM_COMPUTE_ERROR("Not supported");
-                    break;
-            }
-            break;
-        }
-        default:
-            break;
-    }
-    ARM_COMPUTE_ERROR_ON_MSG_VAR(num_elems_processed_per_iteration == 0, "Conversion from %s to %s not supported",
-                                 string_from_format(input->info()->format()).c_str(),
-                                 string_from_format(output->info()->format()).c_str());
-
-    std::stringstream kernel_name;
-
-    kernel_name << string_from_format(input->info()->format());
-    kernel_name << "_to_";
-    kernel_name << string_from_format(output->info()->format());
-    kernel_name << "_bt709";
-
-    _multi_input  = input;
-    _multi_output = output;
-
-    // Create kernel
-    bool has_two_input_planars  = (input->info()->format() == Format::NV12) || (input->info()->format() == Format::NV21);
-    bool has_two_output_planars = (output->info()->format() == Format::NV12) || (output->info()->format() == Format::NV21);
-
-    float sub_sampling_input  = (has_two_input_planars || (input->info()->format() == Format::IYUV)) ? 0.5f : 1;
-    float sub_sampling_output = (has_two_output_planars || (output->info()->format() == Format::IYUV)) ? 0.5f : 1;
-
-    _kernel = create_kernel(compile_context, kernel_name.str());
-
-    Window win = calculate_max_window(*input->cl_plane(0)->info(), Steps(num_elems_processed_per_iteration));
-    win.set_dimension_step(Window::DimY, 2);
-
-    AccessWindowHorizontal input_plane0_access(input->plane(0)->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowRectangle  input_plane1_access(input->plane(1)->info(), 0, 0, num_elems_processed_per_iteration, 1,
-                                               sub_sampling_input, sub_sampling_input);
-    AccessWindowRectangle input_plane2_access(has_two_input_planars ? nullptr : input->plane(2)->info(), 0, 0, num_elems_processed_per_iteration, 1,
-                                              sub_sampling_input, sub_sampling_input);
-    AccessWindowHorizontal output_plane0_access(output->plane(0)->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowRectangle  output_plane1_access(output->plane(1)->info(), 0, 0, num_elems_processed_per_iteration, 1, sub_sampling_output, sub_sampling_output);
-    AccessWindowRectangle  output_plane2_access(has_two_output_planars ? nullptr : output->plane(2)->info(), 0, 0,
-                                                num_elems_processed_per_iteration, 1, sub_sampling_output, sub_sampling_output);
-
-    update_window_and_padding(win,
-                              input_plane0_access, input_plane1_access, input_plane2_access,
-                              output_plane0_access, output_plane1_access, output_plane2_access);
-
-    ValidRegion intersect_region = intersect_valid_regions(input->plane(0)->info()->valid_region(), input->plane(1)->info()->valid_region(),
-                                                           input->plane(2)->info()->valid_region());
-    output_plane0_access.set_valid_region(win, ValidRegion(intersect_region.anchor, output->plane(0)->info()->tensor_shape()));
-    output_plane1_access.set_valid_region(win, ValidRegion(intersect_region.anchor, output->plane(1)->info()->tensor_shape()));
-    output_plane2_access.set_valid_region(win, ValidRegion(intersect_region.anchor, output->plane(2)->info()->tensor_shape()));
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name.str();
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->plane(0)->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->plane(0)->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->plane(0)->info()->dimension(1));
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->plane(1)->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->plane(1)->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->plane(1)->info()->dimension(1));
-}
-
-void CLColorConvertKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-
-    if(nullptr != _input && nullptr != _output)
-    {
-        do
-        {
-            unsigned int idx = 0;
-            add_2D_tensor_argument(idx, _input, slice);
-            add_2D_tensor_argument(idx, _output, slice);
-            enqueue(queue, *this, slice, lws_hint());
-        }
-        while(window.slide_window_slice_2D(slice));
-    }
-    else if(nullptr != _input && nullptr != _multi_output)
-    {
-        Format format = _multi_output->info()->format();
-        do
-        {
-            Window win_uv(slice);
-
-            if((Format::NV12 == format) || (Format::NV21 == format) || (Format::IYUV == format))
-            {
-                win_uv.set(Window::DimX, Window::Dimension(win_uv.x().start() / 2, win_uv.x().end() / 2, win_uv.x().step() / 2));
-                win_uv.set(Window::DimY, Window::Dimension(win_uv.y().start() / 2, win_uv.y().end() / 2, 1));
-            }
-            unsigned int idx = 0;
-            add_2D_tensor_argument(idx, _input, slice);
-            add_2D_tensor_argument(idx, _multi_output->cl_plane(0), slice);
-            for(int i = 1; i < 3 && (0 != _multi_output->cl_plane(i)->info()->num_dimensions()); ++i)
-            {
-                add_2D_tensor_argument(idx, _multi_output->cl_plane(i), win_uv);
-            }
-            enqueue(queue, *this, slice, lws_hint());
-        }
-        while(window.slide_window_slice_2D(slice));
-    }
-    else if(nullptr != _multi_input && nullptr != _output)
-    {
-        Format format = _multi_input->info()->format();
-        do
-        {
-            Window win_uv(slice);
-
-            if((Format::NV12 == format) || (Format::NV21 == format) || (Format::IYUV == format))
-            {
-                win_uv.set(Window::DimX, Window::Dimension(win_uv.x().start() / 2, win_uv.x().end() / 2, win_uv.x().step() / 2));
-                win_uv.set(Window::DimY, Window::Dimension(win_uv.y().start() / 2, win_uv.y().end() / 2, 1));
-            }
-
-            unsigned int idx = 0;
-            add_2D_tensor_argument(idx, _multi_input->cl_plane(0), slice);
-
-            for(int i = 1; i < 3 && (0 != _multi_input->cl_plane(i)->info()->num_dimensions()); ++i)
-            {
-                add_2D_tensor_argument(idx, _multi_input->cl_plane(i), win_uv);
-            }
-            add_2D_tensor_argument(idx, _output, slice);
-            enqueue(queue, *this, slice, lws_hint());
-        }
-        while(window.slide_window_slice_2D(slice));
-    }
-    else if(nullptr != _multi_input && nullptr != _multi_output)
-    {
-        Format in_format  = _multi_input->info()->format();
-        Format out_format = _multi_output->info()->format();
-        do
-        {
-            Window win_in_uv(slice);
-            if((Format::NV12 == in_format) || (Format::NV21 == in_format) || (Format::IYUV == in_format))
-            {
-                win_in_uv.set(Window::DimX, Window::Dimension(win_in_uv.x().start() / 2,
-                                                              win_in_uv.x().end() / 2, win_in_uv.x().step() / 2));
-                win_in_uv.set(Window::DimY, Window::Dimension(win_in_uv.y().start() / 2, win_in_uv.y().end() / 2, 1));
-            }
-            unsigned int idx = 0;
-            add_2D_tensor_argument(idx, _multi_input->cl_plane(0), slice);
-            for(int i = 1; i < 3 && (0 != _multi_input->cl_plane(i)->info()->num_dimensions()); ++i)
-            {
-                add_2D_tensor_argument(idx, _multi_input->cl_plane(i), win_in_uv);
-            }
-
-            Window win_out_uv(slice);
-            if((Format::NV12 == out_format) || (Format::NV21 == out_format) || (Format::IYUV == out_format))
-            {
-                win_out_uv.set(Window::DimX, Window::Dimension(win_out_uv.x().start() / 2,
-                                                               win_out_uv.x().end() / 2, win_out_uv.x().step() / 2));
-                win_out_uv.set(Window::DimY, Window::Dimension(win_out_uv.y().start() / 2, win_out_uv.y().end() / 2, 1));
-            }
-
-            add_2D_tensor_argument(idx, _multi_output->cl_plane(0), slice);
-            for(int i = 1; i < 3 && (0 != _multi_output->cl_plane(i)->info()->num_dimensions()); ++i)
-            {
-                add_2D_tensor_argument(idx, _multi_output->cl_plane(i), win_out_uv);
-            }
-            enqueue(queue, *this, slice, lws_hint());
-        }
-        while(window.slide_window_slice_2D(slice));
-    }
-    else
-    {
-        ARM_COMPUTE_ERROR("Not supported");
-    }
-}
diff --git a/src/core/CL/kernels/CLColorConvertKernel.h b/src/core/CL/kernels/CLColorConvertKernel.h
deleted file mode 100644
index 0f082914cd..0000000000
--- a/src/core/CL/kernels/CLColorConvertKernel.h
+++ /dev/null
@@ -1,121 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLCOLORCONVERTKERNEL_H
-#define ARM_COMPUTE_CLCOLORCONVERTKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLMultiImage;
-class ICLTensor;
-using ICLImage = ICLTensor;
-
-/** Interface for the color convert kernel.
- *
- */
-class CLColorConvertKernel : public ICLKernel
-{
-public:
-    /** Default constructor. */
-    CLColorConvertKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLColorConvertKernel(const CLColorConvertKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLColorConvertKernel &operator=(const CLColorConvertKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLColorConvertKernel(CLColorConvertKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLColorConvertKernel &operator=(CLColorConvertKernel &&) = default;
-    /** Default destructor. */
-    ~CLColorConvertKernel() = default;
-
-    /** Set the input and output of the kernel
-     *
-     * @param[in]  input  Source tensor. Formats supported: RGBA8888/UYVY422/YUYV422/RGB888
-     * @param[out] output Destination tensor. Formats supported: RGB888 (if the formats of @p input are RGBA8888/UYVY422/YUYV422),
-     *                                                          RGBA8888 (if the formats of @p input are UYVY422/YUYV422/RGB888/),
-     *                                                          U8 (if the formats of @p input is RGB888)
-     */
-    void configure(const ICLTensor *input, ICLTensor *output);
-    /** Set the input and output of the kernel
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Source tensor. Formats supported: RGBA8888/UYVY422/YUYV422/RGB888
-     * @param[out] output          Destination tensor. Formats supported: RGB888 (if the formats of @p input are RGBA8888/UYVY422/YUYV422),
-     *                                                          RGBA8888 (if the formats of @p input are UYVY422/YUYV422/RGB888/),
-     *                                                          U8 (if the formats of @p input is RGB888)
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output);
-    /** Set the input and output of the kernel
-     *
-     * @param[in]  input  Multi-planar source image. Formats supported: NV12/NV21/IYUV
-     * @param[out] output Single-planar destination image. Formats supported: RGB888/RGBA8888
-     */
-    void configure(const ICLMultiImage *input, ICLImage *output);
-    /** Set the input and output of the kernel
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Multi-planar source image. Formats supported: NV12/NV21/IYUV
-     * @param[out] output          Single-planar destination image. Formats supported: RGB888/RGBA8888
-     */
-    void configure(const CLCompileContext &compile_context, const ICLMultiImage *input, ICLImage *output);
-    /** Set the input and output of the kernel
-     *
-     * @param[in]  input  Single-planar source image. Formats supported: RGB888/RGBA8888/UYVY422/YUYV422
-     * @param[out] output Multi-planar destination image. Formats supported: NV12/IYUV/YUV444 (if the formats of @p input are RGB888/RGB8888)
-     */
-    void configure(const ICLImage *input, ICLMultiImage *output);
-    /** Set the input and output of the kernel
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Single-planar source image. Formats supported: RGB888/RGBA8888/UYVY422/YUYV422
-     * @param[out] output          Multi-planar destination image. Formats supported: NV12/IYUV/YUV444 (if the formats of @p input are RGB888/RGB8888)
-     */
-    void configure(const CLCompileContext &compile_context, const ICLImage *input, ICLMultiImage *output);
-    /** Set the input and output of the kernel
-     *
-     * @param[in]  input  Multi-planar source image. Formats supported: NV12/NV21/IYUV
-     * @param[out] output Multi-planar destination image. Formats supported: YUV444/IYUV (if the formats of @p input are NV12/NV21)/NV12 (if the format of  @p input is IYUV)
-     */
-    void configure(const ICLMultiImage *input, ICLMultiImage *output);
-    /** Set the input and output of the kernel
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Multi-planar source image. Formats supported: NV12/NV21/IYUV
-     * @param[out] output          Multi-planar destination image. Formats supported: YUV444/IYUV (if the formats of @p input are NV12/NV21)/NV12 (if the format of  @p input is IYUV)
-     */
-    void configure(const CLCompileContext &compile_context, const ICLMultiImage *input, ICLMultiImage *output);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor     *_input;        /*pointer to single planar tensor input */
-    ICLTensor           *_output;       /*pointer to single planar tensor output */
-    const ICLMultiImage *_multi_input;  /*pointer to multi-planar input */
-    ICLMultiImage       *_multi_output; /*pointer to multi-planar output */
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLCOLORCONVERTKERNEL_H */
diff --git a/src/core/CL/kernels/CLConvolutionKernel.cpp b/src/core/CL/kernels/CLConvolutionKernel.cpp
deleted file mode 100644
index 21f1047cc6..0000000000
--- a/src/core/CL/kernels/CLConvolutionKernel.cpp
+++ /dev/null
@@ -1,392 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLConvolutionKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Error.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/CL/ICLKernel.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <set>
-#include <sstream>
-#include <string>
-
-namespace arm_compute
-{
-namespace
-{
-constexpr unsigned int max_matrix_size = 81;
-} // namespace
-
-/****************************************************************************************\
- *                                 Square Convolution                                *
-\****************************************************************************************/
-
-template <unsigned int matrix_size>
-BorderSize             CLConvolutionKernel<matrix_size>::border_size() const
-{
-    return BorderSize(matrix_size / 2);
-}
-
-template <unsigned int matrix_size>
-void CLConvolutionKernel<matrix_size>::configure(const ICLTensor *input, ICLTensor *output, const int16_t *conv, uint32_t scale, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, conv, scale, border_undefined);
-}
-
-template <unsigned int matrix_size>
-void CLConvolutionKernel<matrix_size>::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const int16_t *conv, uint32_t scale, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8, DataType::S16);
-    ARM_COMPUTE_ERROR_ON(conv == nullptr);
-
-    _input  = input;
-    _output = output;
-
-    std::stringstream kernel_name;
-    CLBuildOptions    build_opts;
-    kernel_name << "convolution" << matrix_size << "x" << matrix_size << "_static";
-
-    if(scale == 0)
-    {
-        scale = calculate_matrix_scale(conv, matrix_size);
-    }
-
-    for(unsigned int i = 0; i < matrix_size * matrix_size; i++)
-    {
-        std::stringstream mat_str;
-        mat_str << "-DMAT" << i << "=" << conv[i];
-        build_opts.add_option(mat_str.str());
-    }
-
-    build_opts.add_option("-DSCALE=" + support::cpp11::to_string(scale));
-
-    DataType data_type = data_type_for_convolution_matrix(conv, matrix_size * matrix_size);
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(data_type));
-
-    std::stringstream out_type;
-    out_type << "-DDATA_TYPE_OUT=" << get_cl_type_from_data_type(output->info()->data_type());
-    build_opts.add_option(out_type.str());
-
-    _kernel = create_kernel(compile_context, kernel_name.str(), build_opts.options());
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    constexpr unsigned int num_rows_read_per_iteration       = matrix_size;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-}
-
-/****************************************************************************************\
- *                                 Separable Convolution                                *
-\****************************************************************************************/
-template <unsigned int matrix_size>
-CLSeparableConvolutionHorKernel<matrix_size>::CLSeparableConvolutionHorKernel()
-    : _border_size(0)
-{
-}
-
-template <unsigned int matrix_size>
-BorderSize             CLSeparableConvolutionHorKernel<matrix_size>::border_size() const
-{
-    return _border_size;
-}
-
-template <unsigned int matrix_size>
-void CLSeparableConvolutionHorKernel<matrix_size>::configure(const ICLTensor *input, ICLTensor *output, const int16_t *conv, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, conv, border_undefined);
-}
-
-template <unsigned int matrix_size>
-void CLSeparableConvolutionHorKernel<matrix_size>::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const int16_t *conv, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U16, DataType::S16, DataType::S32);
-
-    ARM_COMPUTE_ERROR_ON((matrix_size != 5) && (matrix_size != 7) && (matrix_size != 9));
-
-    _input       = input;
-    _output      = output;
-    _border_size = BorderSize(border_undefined ? 0 : matrix_size / 2, matrix_size / 2);
-
-    // Set build options
-    std::set<std::string> build_opts;
-
-    std::array<int16_t, matrix_size *matrix_size> mat = { 0 };
-    memcpy(mat.data(), conv, matrix_size * sizeof(int16_t));
-
-    for(unsigned int j = 0; j < matrix_size * matrix_size; j++)
-    {
-        build_opts.insert("-DMAT" + support::cpp11::to_string(j) + "=" + support::cpp11::to_string(mat[j]));
-    }
-
-    build_opts.insert("-DSCALE=0");
-
-    build_opts.insert("-DDATA_TYPE=" + get_cl_type_from_data_type(output->info()->data_type()));
-
-    // Create kernel
-    const std::string kernel_name = "convolution_separable1x" + support::cpp11::to_string(matrix_size) + "_static";
-    _kernel                       = create_kernel(compile_context, kernel_name, build_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-
-    Window win = calculate_max_window_horizontal(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowHorizontal input_access(input->info(), -border_size().left, num_elems_read_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(border_undefined);
-}
-
-template <unsigned int matrix_size>
-BorderSize             CLSeparableConvolutionVertKernel<matrix_size>::border_size() const
-{
-    return BorderSize{ matrix_size / 2, 0 };
-}
-
-template <unsigned int matrix_size>
-void CLSeparableConvolutionVertKernel<matrix_size>::configure(const ICLTensor *input, ICLTensor *output,
-                                                              const int16_t *conv, uint32_t scale, bool border_undefined, DataType data_type)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, conv, scale, border_undefined, data_type);
-}
-
-template <unsigned int matrix_size>
-void CLSeparableConvolutionVertKernel<matrix_size>::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output,
-                                                              const int16_t *conv, uint32_t scale, bool border_undefined, DataType data_type)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U16, DataType::S16, DataType::S32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8, DataType::S16);
-    ARM_COMPUTE_ERROR_ON((matrix_size != 5) && (matrix_size != 7) && (matrix_size != 9));
-    ARM_COMPUTE_ERROR_ON(scale == 0);
-
-    _input  = input;
-    _output = output;
-
-    std::set<std::string> build_opts;
-
-    std::array<int16_t, matrix_size *matrix_size> mat = { 0 };
-    memcpy(mat.data() + matrix_size, conv, matrix_size * sizeof(int16_t));
-
-    for(unsigned int j = 0; j < matrix_size * matrix_size; j++)
-    {
-        build_opts.insert("-DMAT" + support::cpp11::to_string(j) + "=" + support::cpp11::to_string(mat[j]));
-    }
-
-    build_opts.insert("-DSCALE=" + support::cpp11::to_string(scale));
-
-    build_opts.insert("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
-
-    build_opts.insert("-DCOMPUTE_TYPE=" + get_cl_type_from_data_type(data_type));
-
-    std::stringstream out_type;
-    out_type << "-DDATA_TYPE_OUT=" << get_cl_type_from_data_type(output->info()->data_type());
-    build_opts.insert(out_type.str());
-
-    // Create kernel
-    const std::string kernel_name = "convolution_separable" + support::cpp11::to_string(matrix_size) + "x1_static";
-    _kernel                       = create_kernel(compile_context, kernel_name, build_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 8;
-    constexpr unsigned int num_rows_read_per_iteration       = matrix_size;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_access(input->info(), 0, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(data_type));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(border_undefined);
-}
-
-/****************************************************************************************\
- *                                 Rectangle Convolution                                *
-\****************************************************************************************/
-
-CLConvolutionRectangleKernel::CLConvolutionRectangleKernel()
-    : _border_size(0), _input(nullptr), _output(nullptr)
-{
-}
-
-BorderSize CLConvolutionRectangleKernel::border_size() const
-{
-    return _border_size;
-}
-
-void CLConvolutionRectangleKernel::configure(const ICLTensor *input, ICLTensor *output, const int16_t *conv, uint32_t width, uint32_t height, uint32_t scale, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, conv, width, height, scale, border_undefined);
-}
-
-void CLConvolutionRectangleKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const int16_t *conv, uint32_t width, uint32_t height, uint32_t scale,
-                                             bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8, DataType::S16);
-    ARM_COMPUTE_ERROR_ON(nullptr == conv);
-    ARM_COMPUTE_ERROR_ON(3 != width && 5 != width && 7 != width && 9 != width);
-    ARM_COMPUTE_ERROR_ON(3 != height && 5 != height && 7 != height && 9 != height);
-    ARM_COMPUTE_ERROR_ON(0 == scale);
-
-    _input       = input;
-    _output      = output;
-    _border_size = BorderSize(height / 2, width / 2);
-
-    std::set<std::string> options;
-
-    std::stringstream output_type;
-    output_type << "-DDATA_TYPE_OUT=" << get_cl_type_from_data_type(output->info()->data_type());
-    options.insert(output_type.str());
-
-    uint32_t matrix_size = width * height;
-
-    std::array<int16_t, max_matrix_size> mat = { 0 };
-
-    memcpy(mat.data(), conv, matrix_size * sizeof(int16_t));
-
-    for(unsigned int j = 0; j < max_matrix_size; j++)
-    {
-        options.insert("-DMAT" + support::cpp11::to_string(j) + "=" + support::cpp11::to_string(mat[j]));
-    }
-
-    options.insert("-DSCALE=" + support::cpp11::to_string(scale));
-
-    DataType data_type = data_type_for_convolution_matrix(conv, matrix_size);
-    options.insert("-DDATA_TYPE=" + get_cl_type_from_data_type(data_type));
-
-    options.insert("-DMATRIX_WIDTH=" + support::cpp11::to_string(width));
-    options.insert("-DMATRIX_HEIGHT=" + support::cpp11::to_string(height));
-
-    _kernel = create_kernel(compile_context, "convolution_rectangle", options);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-    const unsigned int     num_rows_read_per_iteration       = height;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-}
-
-void CLConvolutionRectangleKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-        add_2D_tensor_argument(idx, _output, slice);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
-
-template class arm_compute::CLConvolutionKernel<3>;
-template class arm_compute::CLConvolutionKernel<5>;
-template class arm_compute::CLConvolutionKernel<7>;
-template class arm_compute::CLConvolutionKernel<9>;
-template class arm_compute::CLSeparableConvolutionVertKernel<5>;
-template class arm_compute::CLSeparableConvolutionVertKernel<7>;
-template class arm_compute::CLSeparableConvolutionVertKernel<9>;
-template class arm_compute::CLSeparableConvolutionHorKernel<5>;
-template class arm_compute::CLSeparableConvolutionHorKernel<7>;
-template class arm_compute::CLSeparableConvolutionHorKernel<9>;
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLConvolutionKernel.h b/src/core/CL/kernels/CLConvolutionKernel.h
deleted file mode 100644
index 33e73caf11..0000000000
--- a/src/core/CL/kernels/CLConvolutionKernel.h
+++ /dev/null
@@ -1,224 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLCONVOLUTIONKERNEL_H
-#define ARM_COMPUTE_CLCONVOLUTIONKERNEL_H
-
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-#include <cstdint>
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/****************************************************************************************\
- *                                    Square Convolution                                *
-\****************************************************************************************/
-
-/** Interface for the kernel to run an arbitrary size convolution on a tensor. (Currently supports 3x3, 5x5, 7x7 and 9x9).
- * The client can supply a convolution matrix \f$ C_{m,n} \f$.
- * @f{eqnarray}{
- *  k_0 &=& \frac{m}{2}  \\
- *  l_0 &=& \frac{n}{2}  \\
- *  sum &=& \sum_{k=0,l=0}^{k=m-1,l=n-1} input(x+k-k_0, y+l-l_0) C_{k,l}
- *  @f}
- *
- * @note The above equation for this function is similar to the default OpenCV Filter2D function,
- *       which actually computes a correlation and not a convolution.
- *       In case of a real convolution the convolution matrix should be flipped both horizontally and vertically.
- */
-template <unsigned int matrix_size>
-class CLConvolutionKernel : public ICLSimple2DKernel
-{
-public:
-    /** Initialise the kernel's input, output and border mode.
-     *
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output           Destination tensor, Data types supported: U8, S16.
-     * @param[in]  conv             Convolution matrix to apply to the input tensor.
-     * @param[in]  scale            Scale of the convolution matrix. If 0 is passed, it will be set to the sum of the coefficients of the convolution or 1 if they add up to 0.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, const int16_t *conv, uint32_t scale, bool border_undefined);
-    /** Initialise the kernel's input, output and border mode.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output           Destination tensor, Data types supported: U8, S16.
-     * @param[in]  conv             Convolution matrix to apply to the input tensor.
-     * @param[in]  scale            Scale of the convolution matrix. If 0 is passed, it will be set to the sum of the coefficients of the convolution or 1 if they add up to 0.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const int16_t *conv, uint32_t scale, bool border_undefined);
-
-    // Inherited methods overridden:
-    BorderSize border_size() const override;
-};
-
-/** Interface for the kernel which applies a 3x3 convolution to a tensor. */
-using CLConvolution3x3Kernel = CLConvolutionKernel<3>;
-/** Interface for the kernel which applies a 5x5 convolution to a tensor. */
-using CLConvolution5x5Kernel = CLConvolutionKernel<5>;
-/** Interface for the kernel which applies a 7x7 convolution to a tensor. */
-using CLConvolution7x7Kernel = CLConvolutionKernel<7>;
-/** Interface for the kernel which applies a 9x9 convolution to a tensor. */
-using CLConvolution9x9Kernel = CLConvolutionKernel<9>;
-
-/****************************************************************************************\
- *                              Separable Square Convolution                            *
-\****************************************************************************************/
-
-/** Kernel for the Horizontal pass of a Separable Convolution. Currently support 5x5, 7x7, 9x9 */
-template <unsigned int matrix_size>
-class CLSeparableConvolutionHorKernel : public ICLSimple2DKernel
-{
-public:
-    /** Default Constructor */
-    CLSeparableConvolutionHorKernel();
-    /** Initialise the kernel's input, output and border mode.
-     *
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output           Destination tensor, Data types supported: S16.
-     * @param[in]  conv             Convolution matrix to apply to the input tensor.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, const int16_t *conv, bool border_undefined);
-    /** Initialise the kernel's input, output and border mode.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output           Destination tensor, Data types supported: U16/S16/S32.
-     * @param[in]  conv             Convolution matrix to apply to the input tensor.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const int16_t *conv, bool border_undefined);
-
-    // Inherited methods overridden:
-    BorderSize border_size() const override;
-
-private:
-    BorderSize _border_size; /**< Border size */
-};
-
-/** Interface for the kernel which applies a horizontal pass of 5x5 convolution to a tensor. */
-using CLSeparableConvolution5x5HorKernel = CLSeparableConvolutionHorKernel<5>;
-/** Interface for the kernel which applies a horizontal pass of 7x7 convolution to a tensor. */
-using CLSeparableConvolution7x7HorKernel = CLSeparableConvolutionHorKernel<7>;
-/** Interface for the kernel which applies a horizontal pass of 9x9 convolution to a tensor. */
-using CLSeparableConvolution9x9HorKernel = CLSeparableConvolutionHorKernel<9>;
-
-/** Kernel for the Vertical pass of a Separable Convolution. Currently supports 5x5, 7x7, 9x9 */
-template <unsigned int matrix_size>
-class CLSeparableConvolutionVertKernel : public ICLSimple2DKernel
-{
-public:
-    /** Initialise the kernel's input, output and border mode.
-     *
-     * @param[in]  input            Source tensor. Data types supported: U16/S16/S32.
-     * @param[out] output           Destination tensor, Data types supported: U8, S16.
-     * @param[in]  conv             Convolution matrix to apply to the input tensor.
-     * @param[in]  scale            Scale of the convolution matrix.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     * @param[in]  data_type        Data type to use for intermeidate result. @sa data_type_for_convolution
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, const int16_t *conv, uint32_t scale, bool border_undefined, DataType data_type = DataType::S32);
-    /** Initialise the kernel's input, output and border mode.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            Source tensor. Data types supported: U16/S16/S32.
-     * @param[out] output           Destination tensor, Data types supported: U8, S16.
-     * @param[in]  conv             Convolution matrix to apply to the input tensor.
-     * @param[in]  scale            Scale of the convolution matrix.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     * @param[in]  data_type        Data type to use for intermeidate result. @sa data_type_for_convolution
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const int16_t *conv, uint32_t scale, bool border_undefined, DataType data_type = DataType::S32);
-
-    // Inherited methods overridden:
-    BorderSize border_size() const override;
-};
-
-/** Interface for the kernel which applies a vertical pass of 5x5 convolution to a tensor. */
-using CLSeparableConvolution5x5VertKernel = CLSeparableConvolutionVertKernel<5>;
-/** Interface for the kernel which applies a vertical pass of 7x7 convolution to a tensor. */
-using CLSeparableConvolution7x7VertKernel = CLSeparableConvolutionVertKernel<7>;
-/** Interface for the kernel which applies a vertical pass of 9x9 convolution to a tensor. */
-using CLSeparableConvolution9x9VertKernel = CLSeparableConvolutionVertKernel<9>;
-
-/****************************************************************************************\
- *                                 Rectangle Convolution                                *
-\****************************************************************************************/
-
-/** Kernel for the running convolution on a rectangle matrix.
- *
- * @note Supports combinations of 3,5,7 and 9.
- */
-class CLConvolutionRectangleKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLConvolutionRectangleKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLConvolutionRectangleKernel(const CLConvolutionRectangleKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLConvolutionRectangleKernel &operator=(const CLConvolutionRectangleKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLConvolutionRectangleKernel(CLConvolutionRectangleKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLConvolutionRectangleKernel &operator=(CLConvolutionRectangleKernel &&) = default;
-    /** Initialise the kernel's input, output and border mode.
-     *
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output           Destination tensor, Data types supported: U8, S16.
-     * @param[in]  conv             Convolution matrix to apply to the input tensor.
-     * @param[in]  width            Width of convolution matrix (Number of columns)
-     * @param[in]  height           Height of convolution matrix (Number of rows)
-     * @param[in]  scale            Scale of the convolution matrix. If 0 is passed, it will be set to the sum of the coefficients of the convolution or 1 if they add up to 0.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, const int16_t *conv, uint32_t width, uint32_t height, uint32_t scale, bool border_undefined);
-    /** Initialise the kernel's input, output and border mode.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output           Destination tensor, Data types supported: U8, S16.
-     * @param[in]  conv             Convolution matrix to apply to the input tensor.
-     * @param[in]  width            Width of convolution matrix (Number of columns)
-     * @param[in]  height           Height of convolution matrix (Number of rows)
-     * @param[in]  scale            Scale of the convolution matrix. If 0 is passed, it will be set to the sum of the coefficients of the convolution or 1 if they add up to 0.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const int16_t *conv, uint32_t width, uint32_t height, uint32_t scale, bool border_undefined);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-    BorderSize border_size() const override;
-
-private:
-    BorderSize       _border_size;
-    const ICLTensor *_input;
-    ICLTensor       *_output;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLCONVOLUTIONKERNEL_H */
diff --git a/src/core/CL/kernels/CLDerivativeKernel.cpp b/src/core/CL/kernels/CLDerivativeKernel.cpp
deleted file mode 100644
index 5ff11362cc..0000000000
--- a/src/core/CL/kernels/CLDerivativeKernel.cpp
+++ /dev/null
@@ -1,155 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLDerivativeKernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <set>
-#include <string>
-
-using namespace arm_compute;
-
-CLDerivativeKernel::CLDerivativeKernel()
-    : _input(nullptr), _output_x(nullptr), _output_y(nullptr), _run_derivative_x(false), _run_derivative_y(false)
-{
-}
-
-BorderSize CLDerivativeKernel::border_size() const
-{
-    return BorderSize(1);
-}
-
-void CLDerivativeKernel::configure(const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output_x, output_y, border_undefined);
-}
-
-void CLDerivativeKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON((output_x == nullptr) && (output_y == nullptr));
-
-    _run_derivative_x = output_x != nullptr;
-    _run_derivative_y = output_y != nullptr;
-
-    if(_run_derivative_x)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_x, 1, DataType::S16);
-    }
-
-    if(_run_derivative_y)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_y, 1, DataType::S16);
-    }
-
-    _input    = input;
-    _output_x = output_x;
-    _output_y = output_y;
-
-    // Set build options
-    std::set<std::string> build_opts;
-
-    if(_run_derivative_x)
-    {
-        build_opts.insert("-DGRAD_X");
-    }
-
-    if(_run_derivative_y)
-    {
-        build_opts.insert("-DGRAD_Y");
-    }
-
-    // Create kernel
-    const std::string kernel_name = std::string("derivative");
-    _kernel                       = create_kernel(compile_context, kernel_name, build_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 16;
-    constexpr unsigned int num_read_rows_per_iteration       = 3;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_access(input->info(), 0, 0, 0, 0);
-    AccessWindowHorizontal output_x_access(output_x == nullptr ? nullptr : output_x->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal output_y_access(output_y == nullptr ? nullptr : output_y->info(), 0, num_elems_processed_per_iteration);
-    if(_run_derivative_x && _run_derivative_y)
-    {
-        // TODO(COMPMID-415) Fix x-access input bug in CL kernel instead of '+2'
-        input_access = AccessWindowRectangle(input->info(), -border_size().left, -border_size().top, num_elems_processed_per_iteration + 2, num_read_rows_per_iteration);
-    }
-    else if(_run_derivative_x)
-    {
-        // TODO(COMPMID-415) Fix x-access input bug in CL kernel instead of '+2'
-        input_access = AccessWindowHorizontal(input->info(), -border_size().left, num_elems_processed_per_iteration + 2);
-    }
-    else if(_run_derivative_y)
-    {
-        input_access = AccessWindowRectangle(input->info(), 0, -border_size().top, num_elems_processed_per_iteration, num_read_rows_per_iteration);
-    }
-
-    update_window_and_padding(win,
-                              input_access,
-                              output_x_access,
-                              output_y_access);
-
-    output_x_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-    output_y_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(border_undefined);
-}
-
-void CLDerivativeKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-        add_2D_tensor_argument_if((_run_derivative_x), idx, _output_x, slice);
-        add_2D_tensor_argument_if((_run_derivative_y), idx, _output_y, slice);
-
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLDerivativeKernel.h b/src/core/CL/kernels/CLDerivativeKernel.h
deleted file mode 100644
index 14dd05d084..0000000000
--- a/src/core/CL/kernels/CLDerivativeKernel.h
+++ /dev/null
@@ -1,83 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLDERIVATIVEKERNEL_H
-#define ARM_COMPUTE_CLDERIVATIVEKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the derivative kernel. */
-class CLDerivativeKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLDerivativeKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLDerivativeKernel(const CLDerivativeKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLDerivativeKernel &operator=(const CLDerivativeKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLDerivativeKernel(CLDerivativeKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLDerivativeKernel &operator=(CLDerivativeKernel &&) = default;
-    /** Default destructor */
-    ~CLDerivativeKernel() = default;
-    /** Initialise the kernel's sources, destination and border
-     *
-     * @note At least one of output_x or output_y must be set
-     *
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S16.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S16.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-    /** Initialise the kernel's sources, destination and border
-     *
-     * @note At least one of output_x or output_y must be set
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S16.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S16.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-    BorderSize border_size() const override;
-
-private:
-    const ICLTensor *_input;            /**< Input tensor */
-    ICLTensor       *_output_x;         /**< Output tensor - Derivate along the X direction */
-    ICLTensor       *_output_y;         /**< Output tensor - Derivate along the Y direction */
-    bool             _run_derivative_x; /**< Do we need to run Derivative X ? */
-    bool             _run_derivative_y; /**< Do we need to run Derivative Y ? */
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLDERIVATIVEKERNEL_H */
diff --git a/src/core/CL/kernels/CLDilateKernel.cpp b/src/core/CL/kernels/CLDilateKernel.cpp
deleted file mode 100644
index cac5bc1c72..0000000000
--- a/src/core/CL/kernels/CLDilateKernel.cpp
+++ /dev/null
@@ -1,70 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLDilateKernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-
-using namespace arm_compute;
-
-BorderSize CLDilateKernel::border_size() const
-{
-    return BorderSize(1);
-}
-
-void CLDilateKernel::configure(const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, border_undefined);
-}
-
-void CLDilateKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8);
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "dilate");
-
-    _input  = input;
-    _output = output;
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    constexpr unsigned int num_rows_read_per_iteration       = 3;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-}
diff --git a/src/core/CL/kernels/CLDilateKernel.h b/src/core/CL/kernels/CLDilateKernel.h
deleted file mode 100644
index 591ec8ccfc..0000000000
--- a/src/core/CL/kernels/CLDilateKernel.h
+++ /dev/null
@@ -1,59 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLDILATEKERNEL_H
-#define ARM_COMPUTE_CLDILATEKERNEL_H
-
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the dilate kernel.
- *
- */
-class CLDilateKernel : public ICLSimple2DKernel
-{
-public:
-    /**Initialise the kernel's input and output.
-     *
-     * @param[in]  input            An input tensor. Data types supported: U8
-     * @param[out] output           The output tensor. Data types supported: U8.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, bool border_undefined);
-    /**Initialise the kernel's input and output.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            An input tensor. Data types supported: U8
-     * @param[out] output           The output tensor. Data types supported: U8.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined);
-
-    // Inherited methods overridden:
-    BorderSize border_size() const override;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLDILATEKERNEL_H */
diff --git a/src/core/CL/kernels/CLErodeKernel.cpp b/src/core/CL/kernels/CLErodeKernel.cpp
deleted file mode 100644
index f6d98a5488..0000000000
--- a/src/core/CL/kernels/CLErodeKernel.cpp
+++ /dev/null
@@ -1,70 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLErodeKernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-
-using namespace arm_compute;
-
-BorderSize CLErodeKernel::border_size() const
-{
-    return BorderSize(1);
-}
-
-void CLErodeKernel::configure(const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, border_undefined);
-}
-
-void CLErodeKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8);
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "erode");
-
-    _input  = input;
-    _output = output;
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    constexpr unsigned int num_rows_read_pes_iteration       = 3;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_pes_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-}
diff --git a/src/core/CL/kernels/CLErodeKernel.h b/src/core/CL/kernels/CLErodeKernel.h
deleted file mode 100644
index 4da97ae358..0000000000
--- a/src/core/CL/kernels/CLErodeKernel.h
+++ /dev/null
@@ -1,59 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLERODEKERNEL_H
-#define ARM_COMPUTE_CLERODEKERNEL_H
-
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the erode kernel.
- *
- */
-class CLErodeKernel : public ICLSimple2DKernel
-{
-public:
-    /**Initialise the kernel's input and output.
-     *
-     * @param[in]  input            An input tensor. Data types supported: U8
-     * @param[out] output           The output tensor. Data types supported: U8.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, bool border_undefined);
-    /**Initialise the kernel's input and output.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            An input tensor. Data types supported: U8
-     * @param[out] output           The output tensor. Data types supported: U8.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined);
-
-    // Inherited methods overridden:
-    BorderSize border_size() const override;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLERODEKERNEL_H */
diff --git a/src/core/CL/kernels/CLFastCornersKernel.cpp b/src/core/CL/kernels/CLFastCornersKernel.cpp
deleted file mode 100644
index 7481fd1c27..0000000000
--- a/src/core/CL/kernels/CLFastCornersKernel.cpp
+++ /dev/null
@@ -1,209 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLFastCornersKernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <set>
-#include <string>
-
-using namespace arm_compute;
-
-CLFastCornersKernel::CLFastCornersKernel()
-    : ICLKernel(), _input(nullptr), _output(nullptr)
-{
-}
-
-BorderSize CLFastCornersKernel::border_size() const
-{
-    return BorderSize(3);
-}
-
-void CLFastCornersKernel::configure(const ICLImage *input, ICLImage *output, float threshold, bool non_max_suppression, BorderMode border_mode)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, threshold, non_max_suppression, border_mode);
-}
-
-void CLFastCornersKernel::configure(const CLCompileContext &compile_context, const ICLImage *input, ICLImage *output, float threshold, bool non_max_suppression, BorderMode border_mode)
-{
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(input);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(output);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_MSG(border_mode != BorderMode::UNDEFINED, "Not implemented");
-
-    _input  = input;
-    _output = output;
-
-    // Create build options
-    std::set<std::string> build_opts;
-
-    if(non_max_suppression)
-    {
-        build_opts.emplace("-DUSE_MAXSUPPRESSION");
-    }
-
-    // Create kernel
-    const std::string kernel_name = std::string("fast_corners");
-    _kernel                       = create_kernel(compile_context, kernel_name, build_opts);
-
-    // Set static kernel arguments
-    unsigned int idx = 2 * num_arguments_per_2D_tensor(); // Skip the input and output parameters
-    _kernel.setArg<cl_float>(idx, static_cast<float>(threshold));
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 1;
-    constexpr unsigned int num_elems_read_per_iteration      = 7;
-    constexpr unsigned int num_rows_read_per_iteration       = 3;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_mode == BorderMode::UNDEFINED, BorderSize(3));
-
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowRectangle  input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region(), border_mode == BorderMode::UNDEFINED, border_size());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(non_max_suppression);
-    _config_id += "_";
-    _config_id += lower_string(string_from_border_mode(border_mode));
-}
-
-void CLFastCornersKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-        add_2D_tensor_argument(idx, _output, slice);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
-
-CLCopyToArrayKernel::CLCopyToArrayKernel()
-    : ICLKernel(), _input(nullptr), _corners(nullptr), _num_buffer(nullptr)
-{
-}
-
-void CLCopyToArrayKernel::configure(const ICLImage *input, bool update_number, ICLKeyPointArray *corners, cl::Buffer *num_buffers)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, update_number, corners, num_buffers);
-}
-
-void CLCopyToArrayKernel::configure(const CLCompileContext &compile_context, const ICLImage *input, bool update_number, ICLKeyPointArray *corners, cl::Buffer *num_buffers)
-{
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(input);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON(corners == nullptr);
-    ARM_COMPUTE_ERROR_ON(num_buffers == nullptr);
-
-    _input      = input;
-    _corners    = corners;
-    _num_buffer = num_buffers;
-
-    std::set<std::string> build_opts;
-
-    if(update_number)
-    {
-        build_opts.emplace("-DUPDATE_NUMBER");
-    }
-
-    // Create kernel
-    const std::string kernel_name = std::string("copy_to_keypoint");
-    _kernel                       = create_kernel(compile_context, kernel_name, build_opts);
-
-    //Get how many pixels skipped in the x dimension in the previous stages
-    unsigned int offset = _input->info()->valid_region().anchor.x();
-
-    // Set static kernel arguments
-    unsigned int idx = num_arguments_per_2D_tensor(); // Skip the input and output parameters
-    _kernel.setArg<unsigned int>(idx++, _corners->max_num_values());
-    _kernel.setArg<cl_uint>(idx++, offset);
-    _kernel.setArg(idx++, *_num_buffer);
-    _kernel.setArg(idx++, _corners->cl_buffer());
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 1;
-    Window                 win                               = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration));
-    update_window_and_padding(win,
-                              AccessWindowHorizontal(input->info(), 0, num_elems_processed_per_iteration));
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-}
-
-void CLCopyToArrayKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    //Initialise the _num_buffer as it used as both input and output
-    static const unsigned int zero_init = 0;
-    queue.enqueueWriteBuffer(*_num_buffer, CL_FALSE, 0, sizeof(unsigned int), &zero_init);
-
-    Window slice = window.first_slice_window_2D();
-
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLFastCornersKernel.h b/src/core/CL/kernels/CLFastCornersKernel.h
deleted file mode 100644
index 0c1b564c2f..0000000000
--- a/src/core/CL/kernels/CLFastCornersKernel.h
+++ /dev/null
@@ -1,133 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLFASTCORNERSKERNEL_H
-#define ARM_COMPUTE_CLFASTCORNERSKERNEL_H
-
-#include "arm_compute/core/CL/ICLArray.h"
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLKernel.h"
-
-#include <cstdint>
-
-namespace cl
-{
-class Buffer;
-}
-
-namespace arm_compute
-{
-class ICLTensor;
-using ICLImage = ICLTensor;
-
-/** CL kernel to perform fast corners */
-class CLFastCornersKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLFastCornersKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLFastCornersKernel(const CLFastCornersKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLFastCornersKernel &operator=(const CLFastCornersKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLFastCornersKernel(CLFastCornersKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLFastCornersKernel &operator=(CLFastCornersKernel &&) = default;
-    /** Default destructor */
-    ~CLFastCornersKernel() = default;
-
-    /** Initialise the kernel.
-     *
-     * @param[in]  input               Source image. Data types supported: U8.
-     * @param[out] output              Output image. Data types supported: U8.
-     * @param[in]  threshold           Threshold on difference between intensity of the central pixel and pixels on Bresenham's circle of radius 3.
-     * @param[in]  non_max_suppression True if non-maxima suppresion is applied, false otherwise.
-     * @param[in]  border_mode         Strategy to use for borders.
-     */
-    void configure(const ICLImage *input, ICLImage *output, float threshold, bool non_max_suppression, BorderMode border_mode);
-    /** Initialise the kernel.
-     *
-     * @param[in]  compile_context     The compile context to be used.
-     * @param[in]  input               Source image. Data types supported: U8.
-     * @param[out] output              Output image. Data types supported: U8.
-     * @param[in]  threshold           Threshold on difference between intensity of the central pixel and pixels on Bresenham's circle of radius 3.
-     * @param[in]  non_max_suppression True if non-maxima suppresion is applied, false otherwise.
-     * @param[in]  border_mode         Strategy to use for borders.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLImage *input, ICLImage *output, float threshold, bool non_max_suppression, BorderMode border_mode);
-
-    // Inherited methods overridden
-    void run(const Window &window, cl::CommandQueue &queue) override;
-    BorderSize border_size() const override;
-
-private:
-    const ICLImage *_input;
-    ICLImage       *_output;
-};
-
-/** CL kernel to copy keypoints information to ICLKeyPointArray and counts the number of key points */
-class CLCopyToArrayKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLCopyToArrayKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLCopyToArrayKernel(const CLCopyToArrayKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLCopyToArrayKernel &operator=(const CLCopyToArrayKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLCopyToArrayKernel(CLCopyToArrayKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLCopyToArrayKernel &operator=(CLCopyToArrayKernel &&) = default;
-    /** Default destructor */
-    ~CLCopyToArrayKernel() = default;
-
-    /** Initialise the kernel.
-     *
-     * @param[in]  input         Source image. Data types supported: U8.
-     * @param[in]  update_number Flag to indicate whether we need to update the number of corners
-     * @param[out] corners       Array of keypoints to store the results.
-     * @param[out] num_buffers   Number of keypoints to store the results.
-     */
-    void configure(const ICLImage *input, bool update_number, ICLKeyPointArray *corners, cl::Buffer *num_buffers);
-    /** Initialise the kernel.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Source image. Data types supported: U8.
-     * @param[in]  update_number   Flag to indicate whether we need to update the number of corners
-     * @param[out] corners         Array of keypoints to store the results.
-     * @param[out] num_buffers     Number of keypoints to store the results.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLImage *input, bool update_number, ICLKeyPointArray *corners, cl::Buffer *num_buffers);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLImage   *_input;      /**< source image */
-    ICLKeyPointArray *_corners;    /**< destination array */
-    cl::Buffer       *_num_buffer; /**< CL memory to record number of key points in the array */
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLFASTCORNERSKERNEL_H */
diff --git a/src/core/CL/kernels/CLGaussian3x3Kernel.cpp b/src/core/CL/kernels/CLGaussian3x3Kernel.cpp
deleted file mode 100644
index 40e9658ab4..0000000000
--- a/src/core/CL/kernels/CLGaussian3x3Kernel.cpp
+++ /dev/null
@@ -1,81 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLGaussian3x3Kernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-
-#include <set>
-#include <string>
-
-using namespace arm_compute;
-
-BorderSize CLGaussian3x3Kernel::border_size() const
-{
-    return BorderSize(1);
-}
-
-void CLGaussian3x3Kernel::configure(const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, border_undefined);
-}
-
-void CLGaussian3x3Kernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8);
-
-    _input  = input;
-    _output = output;
-
-    // Set build options
-    std::set<std::string> build_opts = { "-DMAT0=1", "-DMAT1=2", "-DMAT2=1",
-                                         "-DMAT3=2", "-DMAT4=4", "-DMAT5=2",
-                                         "-DMAT6=1", "-DMAT7=2", "-DMAT8=1",
-                                         "-DSCALE=16", "-DDATA_TYPE_OUT=uchar"
-                                       };
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "convolution3x3_static", build_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    constexpr unsigned int num_rows_read_per_iteration       = 3;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-}
diff --git a/src/core/CL/kernels/CLGaussian3x3Kernel.h b/src/core/CL/kernels/CLGaussian3x3Kernel.h
deleted file mode 100644
index 139b05d44c..0000000000
--- a/src/core/CL/kernels/CLGaussian3x3Kernel.h
+++ /dev/null
@@ -1,59 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLGAUSSIAN3X3KERNEL_H
-#define ARM_COMPUTE_CLGAUSSIAN3X3KERNEL_H
-
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the Gaussian 3x3 filter kernel.
- *
- */
-class CLGaussian3x3Kernel : public ICLSimple2DKernel
-{
-public:
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in]  input            An input tensor. Data types supported: U8
-     * @param[out] output           The output tensor. Data types supported: U8.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, bool border_undefined);
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            An input tensor. Data types supported: U8
-     * @param[out] output           The output tensor. Data types supported: U8.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined);
-
-    // Inherited methods overridden:
-    BorderSize border_size() const override;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLGAUSSIAN3X3KERNEL_H */
diff --git a/src/core/CL/kernels/CLGaussian5x5Kernel.cpp b/src/core/CL/kernels/CLGaussian5x5Kernel.cpp
deleted file mode 100644
index 46a7576154..0000000000
--- a/src/core/CL/kernels/CLGaussian5x5Kernel.cpp
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLGaussian5x5Kernel.h"
-
-#include <cstdint>
-
-using namespace arm_compute;
-
-void CLGaussian5x5HorKernel::configure(const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, border_undefined);
-}
-
-void CLGaussian5x5HorKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    const std::array<int16_t, 5> matrix = { 1, 4, 6, 4, 1 };
-
-    // Set arguments
-    CLSeparableConvolution5x5HorKernel::configure(compile_context, input, output, matrix.data(), border_undefined);
-}
-
-void CLGaussian5x5VertKernel::configure(const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, border_undefined);
-}
-
-void CLGaussian5x5VertKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    const uint32_t scale = 256;
-    const std::array<int16_t, 5> matrix = { 1, 4, 6, 4, 1 };
-
-    // Set arguments
-    CLSeparableConvolution5x5VertKernel::configure(compile_context, input, output, matrix.data(), scale, border_undefined);
-}
diff --git a/src/core/CL/kernels/CLGaussian5x5Kernel.h b/src/core/CL/kernels/CLGaussian5x5Kernel.h
deleted file mode 100644
index 711710b3b3..0000000000
--- a/src/core/CL/kernels/CLGaussian5x5Kernel.h
+++ /dev/null
@@ -1,83 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLGAUSSIAN5X5KERNEL_H
-#define ARM_COMPUTE_CLGAUSSIAN5X5KERNEL_H
-
-#include "src/core/CL/kernels/CLConvolutionKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the kernel to run the horizontal pass of 5x5 Gaussian filter on a tensor. */
-class CLGaussian5x5HorKernel : public CLSeparableConvolution5x5HorKernel
-{
-public:
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output           Destination tensor. Data types supported: S16.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, bool border_undefined);
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output           Destination tensor. Data types supported: S16.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined);
-
-private:
-    //Make the configure method of the parent class private
-    using CLSeparableConvolution5x5HorKernel::configure;
-};
-
-/** Interface for the kernel to run the vertical pass of 5x5 Gaussian filter on a tensor. */
-class CLGaussian5x5VertKernel : public CLSeparableConvolution5x5VertKernel
-{
-public:
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @param[in]  input            Input tensor(output of horizontal pass). Data types supported: S16.
-     * @param[out] output           Destination tensor. Data types supported: U8.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, bool border_undefined);
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            Input tensor(output of horizontal pass). Data types supported: S16.
-     * @param[out] output           Destination tensor. Data types supported: U8.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined);
-
-private:
-    //Make the configure method of the parent class private
-    using CLSeparableConvolution5x5VertKernel::configure;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLGAUSSIAN5X5KERNEL_H */
diff --git a/src/core/CL/kernels/CLGaussianPyramidKernel.cpp b/src/core/CL/kernels/CLGaussianPyramidKernel.cpp
deleted file mode 100644
index 065f7f7e92..0000000000
--- a/src/core/CL/kernels/CLGaussianPyramidKernel.cpp
+++ /dev/null
@@ -1,247 +0,0 @@
-/*
- * Copyright (c) 2017-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLGaussianPyramidKernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-using namespace arm_compute;
-
-CLGaussianPyramidHorKernel::CLGaussianPyramidHorKernel()
-    : _l2_load_offset(0)
-{
-}
-
-BorderSize CLGaussianPyramidHorKernel::border_size() const
-{
-    return BorderSize{ 0, 2 };
-}
-
-void CLGaussianPyramidHorKernel::configure(const ICLTensor *input, ICLTensor *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output);
-}
-
-void CLGaussianPyramidHorKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U16);
-    ARM_COMPUTE_ERROR_ON(input->info()->dimension(1) != output->info()->dimension(1));
-
-    for(size_t i = 2; i < Coordinates::num_max_dimensions; ++i)
-    {
-        ARM_COMPUTE_ERROR_ON(input->info()->dimension(i) != output->info()->dimension(i));
-    }
-
-    _input  = input;
-    _output = output;
-
-    // Create kernel
-    const std::string kernel_name = std::string("gaussian1x5_sub_x");
-    _kernel                       = create_kernel(compile_context, kernel_name);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 16;
-    constexpr unsigned int num_elems_read_per_iteration      = 20;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-    const float            scale_x                           = static_cast<float>(output->info()->dimension(0)) / input->info()->dimension(0);
-
-    Window                 win = calculate_max_window_horizontal(*input->info(), Steps(num_elems_processed_per_iteration));
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_written_per_iteration, scale_x);
-
-    // Sub sampling selects odd pixels (1, 3, 5, ...) for images with even
-    // width and even pixels (0, 2, 4, ...) for images with odd width. (Whether
-    // a pixel is even or odd is determined based on the tensor shape not the
-    // valid region!)
-    // Thus the offset from which the first pixel (L2) for the convolution is
-    // loaded depends on the anchor and shape of the valid region.
-    // In the case of an even shape (= even image width) we need to load L2
-    // from -2 if the anchor is odd and from -1 if the anchor is even. That
-    // makes sure that L2 is always loaded from an odd pixel.
-    // On the other hand, for an odd shape (= odd image width) we need to load
-    // L2 from -1 if the anchor is odd and from -2 if the anchor is even to
-    // achieve the opposite effect.
-    // The condition can be simplified to checking whether anchor + shape is
-    // odd (-2) or even (-1) as only adding an odd and an even number will have
-    // an odd result.
-    _l2_load_offset = -border_size().left;
-
-    if((_input->info()->valid_region().anchor[0] + _input->info()->valid_region().shape[0]) % 2 == 0)
-    {
-        _l2_load_offset += 1;
-    }
-
-    update_window_and_padding(win,
-                              AccessWindowHorizontal(input->info(), _l2_load_offset, num_elems_read_per_iteration),
-                              output_access);
-
-    output->info()->set_valid_region(ValidRegion(Coordinates(), output->info()->tensor_shape()));
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-}
-
-void CLGaussianPyramidHorKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window win_in(window);
-    win_in.shift(Window::DimX, _l2_load_offset);
-
-    //The output is half the width of the input:
-    Window win_out(window);
-    win_out.scale(Window::DimX, 0.5f);
-
-    Window slice_in  = win_in.first_slice_window_2D();
-    Window slice_out = win_out.first_slice_window_2D();
-
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice_in);
-        add_2D_tensor_argument(idx, _output, slice_out);
-        enqueue(queue, *this, slice_out, lws_hint());
-    }
-    while(win_in.slide_window_slice_2D(slice_in) && win_out.slide_window_slice_2D(slice_out));
-}
-
-CLGaussianPyramidVertKernel::CLGaussianPyramidVertKernel()
-    : _t2_load_offset(0)
-{
-}
-
-BorderSize CLGaussianPyramidVertKernel::border_size() const
-{
-    return BorderSize{ 2, 0 };
-}
-
-void CLGaussianPyramidVertKernel::configure(const ICLTensor *input, ICLTensor *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output);
-}
-
-void CLGaussianPyramidVertKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U16);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON(input->info()->dimension(0) != output->info()->dimension(0));
-
-    for(size_t i = 2; i < Coordinates::num_max_dimensions; ++i)
-    {
-        ARM_COMPUTE_ERROR_ON(input->info()->dimension(i) != output->info()->dimension(i));
-    }
-
-    _input  = input;
-    _output = output;
-
-    // Create kernel
-    const std::string kernel_name = std::string("gaussian5x1_sub_y");
-    _kernel                       = create_kernel(compile_context, "gaussian5x1_sub_y");
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_rows_processed_per_iteration  = 2;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 8;
-    constexpr unsigned int num_rows_per_iteration            = 5;
-
-    const float scale_y = static_cast<float>(output->info()->dimension(1)) / input->info()->dimension(1);
-
-    Window                win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration, num_rows_processed_per_iteration));
-    AccessWindowRectangle output_access(output->info(), 0, 0, num_elems_written_per_iteration, num_rows_per_iteration, 1.f, scale_y);
-
-    // Determine whether we need to load even or odd rows. See above for a
-    // detailed explanation.
-    _t2_load_offset = -border_size().top;
-
-    if((_input->info()->valid_region().anchor[1] + _input->info()->valid_region().shape[1]) % 2 == 0)
-    {
-        _t2_load_offset += 1;
-    }
-
-    update_window_and_padding(win,
-                              AccessWindowRectangle(input->info(), 0, _t2_load_offset, num_elems_read_per_iteration, num_rows_per_iteration),
-                              output_access);
-
-    output->info()->set_valid_region(ValidRegion(Coordinates(), output->info()->tensor_shape()));
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-}
-
-void CLGaussianPyramidVertKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
-    ARM_COMPUTE_ERROR_ON(window.x().step() != 8);
-    ARM_COMPUTE_ERROR_ON(window.y().step() % 2);
-
-    Window win_in(window);
-    win_in.shift(Window::DimY, _t2_load_offset);
-
-    Window win_out(window);
-    win_out.scale(Window::DimY, 0.5f);
-
-    Window slice_in  = win_in.first_slice_window_2D();
-    Window slice_out = win_out.first_slice_window_2D();
-
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice_in);
-        add_2D_tensor_argument(idx, _output, slice_out);
-        enqueue(queue, *this, slice_out, lws_hint());
-    }
-    while(win_in.slide_window_slice_2D(slice_in) && win_out.slide_window_slice_2D(slice_out));
-}
diff --git a/src/core/CL/kernels/CLGaussianPyramidKernel.h b/src/core/CL/kernels/CLGaussianPyramidKernel.h
deleted file mode 100644
index a6595440f6..0000000000
--- a/src/core/CL/kernels/CLGaussianPyramidKernel.h
+++ /dev/null
@@ -1,111 +0,0 @@
-/*
- * Copyright (c) 2017-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLGAUSSIANPYRAMIDKERNEL_H
-#define ARM_COMPUTE_CLGAUSSIANPYRAMIDKERNEL_H
-
-#include "src/core/CL/ICLSimpleKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** OpenCL kernel to perform a Gaussian filter and half scaling across width (horizontal pass) */
-class CLGaussianPyramidHorKernel : public ICLSimpleKernel
-{
-public:
-    /** Default constructor */
-    CLGaussianPyramidHorKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGaussianPyramidHorKernel(const CLGaussianPyramidHorKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGaussianPyramidHorKernel &operator=(const CLGaussianPyramidHorKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLGaussianPyramidHorKernel(CLGaussianPyramidHorKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLGaussianPyramidHorKernel &operator=(CLGaussianPyramidHorKernel &&) = default;
-    /** Default destructor */
-    ~CLGaussianPyramidHorKernel() = default;
-
-    /** Initialise the kernel's source, destination and border mode.
-     *
-     * @param[in]  input  Source tensor. Data types supported: U8.
-     * @param[out] output Destination tensor. Output should have half the input width. Data types supported: U16.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output);
-    /** Initialise the kernel's source, destination and border mode.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Source tensor. Data types supported: U8.
-     * @param[out] output          Destination tensor. Output should have half the input width. Data types supported: U16.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-    BorderSize border_size() const override;
-
-private:
-    int _l2_load_offset;
-};
-
-/** OpenCL kernel to perform a Gaussian filter and half scaling across height (vertical pass) */
-class CLGaussianPyramidVertKernel : public ICLSimpleKernel
-{
-public:
-    /** Default constructor */
-    CLGaussianPyramidVertKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGaussianPyramidVertKernel(const CLGaussianPyramidVertKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGaussianPyramidVertKernel &operator=(const CLGaussianPyramidVertKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLGaussianPyramidVertKernel(CLGaussianPyramidVertKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLGaussianPyramidVertKernel &operator=(CLGaussianPyramidVertKernel &&) = default;
-    /** Default destructor */
-    ~CLGaussianPyramidVertKernel() = default;
-
-    /** Initialise the kernel's source, destination and border mode.
-     *
-     * @param[in]  input  Source tensor. Data types supported: U16.
-     * @param[out] output Destination tensor. Output should have half the input height. Data types supported: U8.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output);
-    /** Initialise the kernel's source, destination and border mode.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Source tensor. Data types supported: U16.
-     * @param[out] output          Destination tensor. Output should have half the input height. Data types supported: U8.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-    BorderSize border_size() const override;
-
-private:
-    int _t2_load_offset;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLGAUSSIANPYRAMIDKERNEL_H */
diff --git a/src/core/CL/kernels/CLHOGDescriptorKernel.cpp b/src/core/CL/kernels/CLHOGDescriptorKernel.cpp
deleted file mode 100644
index cd3f1ee216..0000000000
--- a/src/core/CL/kernels/CLHOGDescriptorKernel.cpp
+++ /dev/null
@@ -1,237 +0,0 @@
-/*
- * Copyright (c) 2017-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLHOGDescriptorKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <set>
-#include <sstream>
-#include <string>
-
-using namespace arm_compute;
-
-CLHOGOrientationBinningKernel::CLHOGOrientationBinningKernel()
-    : _input_magnitude(nullptr), _input_phase(nullptr), _output(nullptr), _cell_size()
-{
-}
-
-void CLHOGOrientationBinningKernel::configure(const ICLTensor *input_magnitude, const ICLTensor *input_phase, ICLTensor *output, const HOGInfo *hog_info)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input_magnitude, input_phase, output, hog_info);
-}
-
-void CLHOGOrientationBinningKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input_magnitude, const ICLTensor *input_phase, ICLTensor *output, const HOGInfo *hog_info)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input_magnitude, 1, DataType::S16);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input_phase, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON(hog_info == nullptr);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, hog_info->num_bins(), DataType::F32);
-    ARM_COMPUTE_ERROR_ON(input_magnitude->info()->dimension(Window::DimX) != input_phase->info()->dimension(Window::DimX));
-    ARM_COMPUTE_ERROR_ON(input_magnitude->info()->dimension(Window::DimY) != input_phase->info()->dimension(Window::DimY));
-
-    _input_magnitude = input_magnitude;
-    _input_phase     = input_phase;
-    _output          = output;
-    _cell_size       = hog_info->cell_size();
-
-    float phase_scale = (PhaseType::SIGNED == hog_info->phase_type() ? hog_info->num_bins() / 360.0f : hog_info->num_bins() / 180.0f);
-    phase_scale *= (PhaseType::SIGNED == hog_info->phase_type() ? 360.0f / 255.0f : 1.0f);
-
-    std::stringstream args_str;
-    args_str << "-DCELL_WIDTH=" << hog_info->cell_size().width << " ";
-    args_str << "-DCELL_HEIGHT=" << hog_info->cell_size().height << " ";
-    args_str << "-DNUM_BINS=" << hog_info->num_bins() << " ";
-    args_str << "-DPHASE_SCALE=" << phase_scale << " ";
-
-    // Construct kernel name
-    std::set<std::string> build_opts = {};
-    build_opts.insert(args_str.str());
-
-    // Create kernel
-    const std::string kernel_name = std::string("hog_orientation_binning");
-    _kernel                       = create_kernel(compile_context, kernel_name, build_opts);
-
-    constexpr unsigned int num_elems_processed_per_iteration = 1;
-    constexpr unsigned int num_elems_read_per_iteration      = 1;
-    const unsigned int     num_rows_read_per_iteration       = hog_info->cell_size().height;
-    constexpr unsigned int num_elems_written_per_iteration   = 1;
-
-    // Configure kernel window
-    Window                 win = calculate_max_window(*output->info(), Steps(num_elems_processed_per_iteration));
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win,
-                              AccessWindowRectangle(input_magnitude->info(), 0, 0, num_elems_read_per_iteration, num_rows_read_per_iteration),
-                              AccessWindowRectangle(input_phase->info(), 0, 0, num_elems_read_per_iteration, num_rows_read_per_iteration),
-                              output_access);
-
-    output->info()->set_valid_region(ValidRegion(Coordinates(), output->info()->tensor_shape()));
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input_magnitude->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input_magnitude->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input_magnitude->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-}
-
-void CLHOGOrientationBinningKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        // Compute slice for the magnitude and phase tensors
-        Window slice_mag_phase = window.first_slice_window_2D();
-        slice_mag_phase.set(Window::DimX, Window::Dimension(window.x().start() * _cell_size.width, window.x().start() * _cell_size.width, _cell_size.width));
-        slice_mag_phase.set(Window::DimY, Window::Dimension(window.y().start() * _cell_size.height, window.y().start() * _cell_size.height, _cell_size.height));
-
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input_magnitude, slice_mag_phase);
-        add_2D_tensor_argument(idx, _input_phase, slice_mag_phase);
-        add_2D_tensor_argument(idx, _output, slice);
-
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
-
-CLHOGBlockNormalizationKernel::CLHOGBlockNormalizationKernel()
-    : _input(nullptr), _output(nullptr), _num_cells_per_block_stride()
-{
-}
-
-void CLHOGBlockNormalizationKernel::configure(const ICLTensor *input, ICLTensor *output, const HOGInfo *hog_info)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, hog_info);
-}
-
-void CLHOGBlockNormalizationKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const HOGInfo *hog_info)
-{
-    ARM_COMPUTE_ERROR_ON(hog_info == nullptr);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, hog_info->num_bins(), DataType::F32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_NOT_IN(output, DataType::F32);
-
-    // Number of cells per block
-    const Size2D num_cells_per_block(hog_info->block_size().width / hog_info->cell_size().width,
-                                     hog_info->block_size().height / hog_info->cell_size().height);
-
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, hog_info->num_bins() * num_cells_per_block.area(), DataType::F32);
-
-    // Number of cells per block stride
-    const Size2D num_cells_per_block_stride(hog_info->block_stride().width / hog_info->cell_size().width,
-                                            hog_info->block_stride().height / hog_info->cell_size().height);
-
-    _input                      = input;
-    _output                     = output;
-    _num_cells_per_block_stride = num_cells_per_block_stride;
-
-    std::stringstream args_str;
-    args_str << "-DL2_HYST_THRESHOLD=" << hog_info->l2_hyst_threshold() << " ";
-    args_str << "-DNUM_CELLS_PER_BLOCK_HEIGHT=" << num_cells_per_block.height << " ";
-    args_str << "-DNUM_BINS_PER_BLOCK_X=" << num_cells_per_block.width *hog_info->num_bins() << " ";
-    args_str << "-DNUM_BINS_PER_BLOCK=" << _output->info()->num_channels() << " ";
-    args_str << "-DL2_NORM=" << static_cast<int>(HOGNormType::L2_NORM) << " ";
-    args_str << "-DL1_NORM=" << static_cast<int>(HOGNormType::L1_NORM) << " ";
-    args_str << "-DL2HYS_NORM=" << static_cast<int>(HOGNormType::L2HYS_NORM) << " ";
-    args_str << "-DHOG_NORM_TYPE=" << static_cast<int>(hog_info->normalization_type()) << " ";
-
-    // Construct kernel name
-    std::set<std::string> build_opts = {};
-    build_opts.insert(args_str.str());
-
-    const std::string kernel_name = std::string("hog_block_normalization");
-    _kernel                       = create_kernel(compile_context, kernel_name, build_opts);
-
-    constexpr unsigned int num_elems_processed_per_iteration = 1;
-    constexpr unsigned int num_elems_read_per_iteration      = 1;
-    const unsigned int     num_rows_read_per_iteration       = num_cells_per_block.height;
-    constexpr unsigned int num_elems_written_per_iteration   = 1;
-    const unsigned int     num_rows_written_per_iteration    = num_cells_per_block.height;
-
-    // Configure kernel window
-    Window                win = calculate_max_window(*output->info(), Steps(num_elems_processed_per_iteration));
-    AccessWindowRectangle output_access(output->info(), 0, 0, num_elems_written_per_iteration, num_rows_written_per_iteration);
-
-    update_window_and_padding(win,
-                              AccessWindowRectangle(input->info(), 0, 0, num_elems_read_per_iteration, num_rows_read_per_iteration),
-                              output_access);
-
-    output_access.set_valid_region(win, ValidRegion(Coordinates(), output->info()->tensor_shape()));
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-}
-
-void CLHOGBlockNormalizationKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        // Compute slice for the magnitude and phase tensors
-        Window slice_in = window.first_slice_window_2D();
-        slice_in.set_dimension_step(Window::DimX, _num_cells_per_block_stride.width);
-        slice_in.set_dimension_step(Window::DimY, _num_cells_per_block_stride.height);
-
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice_in);
-        add_2D_tensor_argument(idx, _output, slice);
-
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLHOGDescriptorKernel.h b/src/core/CL/kernels/CLHOGDescriptorKernel.h
deleted file mode 100644
index eee2fa36bc..0000000000
--- a/src/core/CL/kernels/CLHOGDescriptorKernel.h
+++ /dev/null
@@ -1,122 +0,0 @@
-/*
- * Copyright (c) 2017-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLHOGDESCRIPTORKERNEL_H
-#define ARM_COMPUTE_CLHOGDESCRIPTORKERNEL_H
-
-#include "arm_compute/core/IHOG.h"
-#include "arm_compute/core/Size2D.h"
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ITensor;
-
-/** OpenCL kernel to perform HOG Orientation Binning */
-class CLHOGOrientationBinningKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLHOGOrientationBinningKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHOGOrientationBinningKernel(const CLHOGOrientationBinningKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHOGOrientationBinningKernel &operator=(const CLHOGOrientationBinningKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLHOGOrientationBinningKernel(CLHOGOrientationBinningKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLHOGOrientationBinningKernel &operator=(CLHOGOrientationBinningKernel &&) = default;
-    /** Default destructor */
-    ~CLHOGOrientationBinningKernel() = default;
-
-    /**  Initialise the kernel's inputs, output and HOG's metadata
-     *
-     * @param[in]  input_magnitude Input tensor which stores the magnitude of the gradient for each pixel. Data type supported: S16.
-     * @param[in]  input_phase     Input tensor which stores the phase of the gradient for each pixel. Data type supported: U8
-     * @param[out] output          Output tensor which stores the local HOG for each cell. DataType supported: F32. Number of channels supported: equal to the number of histogram bins per cell
-     * @param[in]  hog_info        HOG's metadata
-     */
-    void configure(const ICLTensor *input_magnitude, const ICLTensor *input_phase, ICLTensor *output, const HOGInfo *hog_info);
-    /**  Initialise the kernel's inputs, output and HOG's metadata
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input_magnitude Input tensor which stores the magnitude of the gradient for each pixel. Data type supported: S16.
-     * @param[in]  input_phase     Input tensor which stores the phase of the gradient for each pixel. Data type supported: U8
-     * @param[out] output          Output tensor which stores the local HOG for each cell. DataType supported: F32. Number of channels supported: equal to the number of histogram bins per cell
-     * @param[in]  hog_info        HOG's metadata
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input_magnitude, const ICLTensor *input_phase, ICLTensor *output, const HOGInfo *hog_info);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_input_magnitude;
-    const ICLTensor *_input_phase;
-    ICLTensor       *_output;
-    Size2D           _cell_size;
-};
-
-/** OpenCL kernel to perform HOG block normalization */
-class CLHOGBlockNormalizationKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLHOGBlockNormalizationKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHOGBlockNormalizationKernel(const CLHOGBlockNormalizationKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHOGBlockNormalizationKernel &operator=(const CLHOGBlockNormalizationKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLHOGBlockNormalizationKernel(CLHOGBlockNormalizationKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLHOGBlockNormalizationKernel &operator=(CLHOGBlockNormalizationKernel &&) = default;
-    /** Default destructor */
-    ~CLHOGBlockNormalizationKernel() = default;
-
-    /** Initialise the kernel's input, output and HOG's metadata
-     *
-     * @param[in]  input    Input tensor which stores the local HOG for each cell. Data type supported: F32. Number of channels supported: equal to the number of histogram bins per cell
-     * @param[out] output   Output tensor which stores the normalised blocks. Data type supported: F32. Number of channels supported: equal to the number of histogram bins per block
-     * @param[in]  hog_info HOG's metadata
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, const HOGInfo *hog_info);
-    /** Initialise the kernel's input, output and HOG's metadata
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Input tensor which stores the local HOG for each cell. Data type supported: F32. Number of channels supported: equal to the number of histogram bins per cell
-     * @param[out] output          Output tensor which stores the normalised blocks. Data type supported: F32. Number of channels supported: equal to the number of histogram bins per block
-     * @param[in]  hog_info        HOG's metadata
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const HOGInfo *hog_info);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_input;
-    ICLTensor       *_output;
-    Size2D           _num_cells_per_block_stride;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLHOGDESCRIPTORKERNEL_H */
diff --git a/src/core/CL/kernels/CLHOGDetectorKernel.cpp b/src/core/CL/kernels/CLHOGDetectorKernel.cpp
deleted file mode 100644
index 861155b9a2..0000000000
--- a/src/core/CL/kernels/CLHOGDetectorKernel.cpp
+++ /dev/null
@@ -1,146 +0,0 @@
-/*
- * Copyright (c) 2017-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLHOGDetectorKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLHOG.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-using namespace arm_compute;
-
-CLHOGDetectorKernel::CLHOGDetectorKernel()
-    : _input(nullptr), _detection_windows(), _num_detection_windows(nullptr)
-{
-}
-
-void CLHOGDetectorKernel::configure(const ICLTensor *input, const ICLHOG *hog, ICLDetectionWindowArray *detection_windows, cl::Buffer *num_detection_windows, const Size2D &detection_window_stride,
-                                    float threshold, uint16_t idx_class)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, hog, detection_windows, num_detection_windows, detection_window_stride, threshold, idx_class);
-}
-
-void CLHOGDetectorKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, const ICLHOG *hog, ICLDetectionWindowArray *detection_windows, cl::Buffer *num_detection_windows,
-                                    const Size2D &detection_window_stride,
-                                    float threshold, uint16_t idx_class)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_NOT_IN(input, DataType::F32);
-    ARM_COMPUTE_ERROR_ON(hog == nullptr);
-    ARM_COMPUTE_ERROR_ON(detection_windows == nullptr);
-    ARM_COMPUTE_ERROR_ON(num_detection_windows == nullptr);
-    ARM_COMPUTE_ERROR_ON((detection_window_stride.width % hog->info()->block_stride().width) != 0);
-    ARM_COMPUTE_ERROR_ON((detection_window_stride.height % hog->info()->block_stride().height) != 0);
-
-    const Size2D &detection_window_size = hog->info()->detection_window_size();
-    const Size2D &block_size            = hog->info()->block_size();
-    const Size2D &block_stride          = hog->info()->block_stride();
-
-    _input                 = input;
-    _detection_windows     = detection_windows;
-    _num_detection_windows = num_detection_windows;
-
-    const unsigned int num_bins_per_descriptor_x   = ((detection_window_size.width - block_size.width) / block_stride.width + 1) * input->info()->num_channels();
-    const unsigned int num_blocks_per_descriptor_y = (detection_window_size.height - block_size.height) / block_stride.height + 1;
-
-    ARM_COMPUTE_ERROR_ON((num_bins_per_descriptor_x * num_blocks_per_descriptor_y + 1) != hog->info()->descriptor_size());
-
-    std::stringstream args_str;
-    args_str << "-DNUM_BLOCKS_PER_DESCRIPTOR_Y=" << num_blocks_per_descriptor_y << " ";
-    args_str << "-DNUM_BINS_PER_DESCRIPTOR_X=" << num_bins_per_descriptor_x << " ";
-    args_str << "-DTHRESHOLD=" << threshold << " ";
-    args_str << "-DMAX_NUM_DETECTION_WINDOWS=" << detection_windows->max_num_values() << " ";
-    args_str << "-DIDX_CLASS=" << idx_class << " ";
-    args_str << "-DDETECTION_WINDOW_WIDTH=" << detection_window_size.width << " ";
-    args_str << "-DDETECTION_WINDOW_HEIGHT=" << detection_window_size.height << " ";
-    args_str << "-DDETECTION_WINDOW_STRIDE_WIDTH=" << detection_window_stride.width << " ";
-    args_str << "-DDETECTION_WINDOW_STRIDE_HEIGHT=" << detection_window_stride.height << " ";
-
-    // Construct kernel name
-    std::set<std::string> build_opts = {};
-    build_opts.insert(args_str.str());
-
-    // Create kernel
-    const std::string kernel_name = std::string("hog_detector");
-    _kernel                       = create_kernel(compile_context, kernel_name, build_opts);
-
-    // Set static kernel arguments
-    unsigned int idx = num_arguments_per_2D_tensor(); // Skip the input parameters
-    _kernel.setArg(idx++, hog->cl_buffer());
-    _kernel.setArg(idx++, detection_windows->cl_buffer());
-    _kernel.setArg(idx++, *_num_detection_windows);
-
-    // Get the number of blocks along the x and y directions of the input tensor
-    const ValidRegion &valid_region = input->info()->valid_region();
-    const size_t       num_blocks_x = valid_region.shape[0];
-    const size_t       num_blocks_y = valid_region.shape[1];
-
-    // Get the number of blocks along the x and y directions of the detection window
-    const size_t num_blocks_per_detection_window_x = detection_window_size.width / block_stride.width;
-    const size_t num_blocks_per_detection_window_y = detection_window_size.height / block_stride.height;
-
-    const size_t window_step_x = detection_window_stride.width / block_stride.width;
-    const size_t window_step_y = detection_window_stride.height / block_stride.height;
-
-    // Configure kernel window
-    Window win;
-    win.set(Window::DimX, Window::Dimension(0, floor_to_multiple(num_blocks_x - num_blocks_per_detection_window_x, window_step_x) + window_step_x, window_step_x));
-    win.set(Window::DimY, Window::Dimension(0, floor_to_multiple(num_blocks_y - num_blocks_per_detection_window_y, window_step_y) + window_step_y, window_step_y));
-
-    constexpr unsigned int num_elems_read_per_iteration = 1;
-    const unsigned int     num_rows_read_per_iteration  = num_blocks_per_descriptor_y;
-
-    update_window_and_padding(win, AccessWindowRectangle(input->info(), 0, 0, num_elems_read_per_iteration, num_rows_read_per_iteration));
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-}
-
-void CLHOGDetectorKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLHOGDetectorKernel.h b/src/core/CL/kernels/CLHOGDetectorKernel.h
deleted file mode 100644
index c28e6ebe74..0000000000
--- a/src/core/CL/kernels/CLHOGDetectorKernel.h
+++ /dev/null
@@ -1,96 +0,0 @@
-/*
- * Copyright (c) 2017-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLHOGDETECTORKERNEL_H
-#define ARM_COMPUTE_CLHOGDETECTORKERNEL_H
-
-#include "arm_compute/core/CL/ICLArray.h"
-#include "arm_compute/core/CL/ICLHOG.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "src/core/CL/ICLKernel.h"
-
-namespace cl
-{
-class Buffer;
-}
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** OpenCL kernel to perform HOG detector kernel using linear SVM */
-class CLHOGDetectorKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLHOGDetectorKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHOGDetectorKernel(const CLHOGDetectorKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHOGDetectorKernel &operator=(const CLHOGDetectorKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLHOGDetectorKernel(CLHOGDetectorKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLHOGDetectorKernel &operator=(CLHOGDetectorKernel &&) = default;
-    /** Default destructor */
-    ~CLHOGDetectorKernel() = default;
-
-    /** Initialise the kernel's input, HOG data-object, detection window, the stride of the detection window, the threshold and index of the object to detect
-     *
-     * @param[in]  input                   Input tensor which stores the HOG descriptor obtained with @ref CLHOGOrientationBinningKernel. Data type supported: F32. Number of channels supported: equal to the number of histogram bins per block
-     * @param[in]  hog                     HOG data object used by @ref CLHOGOrientationBinningKernel and  @ref CLHOGBlockNormalizationKernel
-     * @param[out] detection_windows       Array of @ref DetectionWindow. This array stores all the detected objects
-     * @param[in]  num_detection_windows   Number of detected objects
-     * @param[in]  detection_window_stride Distance in pixels between 2 consecutive detection windows in x and y directions.
-     *                                     It must be multiple of the hog->info()->block_stride()
-     * @param[in]  threshold               (Optional) Threshold for the distance between features and SVM classifying plane
-     * @param[in]  idx_class               (Optional) Index of the class used for evaluating which class the detection window belongs to
-     */
-    void configure(const ICLTensor *input, const ICLHOG *hog, ICLDetectionWindowArray *detection_windows, cl::Buffer *num_detection_windows, const Size2D &detection_window_stride, float threshold = 0.0f,
-                   uint16_t idx_class = 0);
-    /** Initialise the kernel's input, HOG data-object, detection window, the stride of the detection window, the threshold and index of the object to detect
-     *
-     * @param[in]  compile_context         The compile context to be used.
-     * @param[in]  input                   Input tensor which stores the HOG descriptor obtained with @ref CLHOGOrientationBinningKernel. Data type supported: F32. Number of channels supported: equal to the number of histogram bins per block
-     * @param[in]  hog                     HOG data object used by @ref CLHOGOrientationBinningKernel and  @ref CLHOGBlockNormalizationKernel
-     * @param[out] detection_windows       Array of @ref DetectionWindow. This array stores all the detected objects
-     * @param[in]  num_detection_windows   Number of detected objects
-     * @param[in]  detection_window_stride Distance in pixels between 2 consecutive detection windows in x and y directions.
-     *                                     It must be multiple of the hog->info()->block_stride()
-     * @param[in]  threshold               (Optional) Threshold for the distance between features and SVM classifying plane
-     * @param[in]  idx_class               (Optional) Index of the class used for evaluating which class the detection window belongs to
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, const ICLHOG *hog, ICLDetectionWindowArray *detection_windows, cl::Buffer *num_detection_windows,
-                   const Size2D &detection_window_stride, float threshold = 0.0f,
-                   uint16_t idx_class = 0);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue);
-
-private:
-    const ICLTensor         *_input;
-    ICLDetectionWindowArray *_detection_windows;
-    cl::Buffer              *_num_detection_windows;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLHOGDETECTORKERNEL_H */
diff --git a/src/core/CL/kernels/CLHarrisCornersKernel.cpp b/src/core/CL/kernels/CLHarrisCornersKernel.cpp
deleted file mode 100644
index cbc056fb77..0000000000
--- a/src/core/CL/kernels/CLHarrisCornersKernel.cpp
+++ /dev/null
@@ -1,149 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLHarrisCornersKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/AccessWindowStatic.h"
-#include "src/core/helpers/AutoConfiguration.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <set>
-#include <sstream>
-#include <string>
-
-using namespace arm_compute;
-
-CLHarrisScoreKernel::CLHarrisScoreKernel()
-    : _input1(nullptr), _input2(nullptr), _output(nullptr), _sensitivity(), _strength_thresh(), _norm_factor(), _border_size(0)
-{
-}
-
-BorderSize CLHarrisScoreKernel::border_size() const
-{
-    return _border_size;
-}
-
-void CLHarrisScoreKernel::configure(const ICLImage *input1, const ICLImage *input2, ICLImage *output,
-                                    int32_t block_size, float norm_factor, float strength_thresh, float sensitivity,
-                                    bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input1, input2, output, block_size, norm_factor, strength_thresh, sensitivity, border_undefined);
-}
-
-void CLHarrisScoreKernel::configure(const CLCompileContext &compile_context, const ICLImage *input1, const ICLImage *input2, ICLImage *output,
-                                    int32_t block_size, float norm_factor, float strength_thresh, float sensitivity,
-                                    bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(input1);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(input2);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(output);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input1, 1, DataType::S16, DataType::S32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input2, 1, DataType::S16, DataType::S32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::F32);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input1, input2);
-    ARM_COMPUTE_ERROR_ON(!(block_size == 3 || block_size == 5 || block_size == 7));
-    ARM_COMPUTE_ERROR_ON(0.0f == norm_factor);
-
-    _input1          = input1;
-    _input2          = input2;
-    _output          = output;
-    _sensitivity     = sensitivity;
-    _strength_thresh = strength_thresh;
-    _norm_factor     = norm_factor;
-    _border_size     = BorderSize(block_size / 2);
-
-    // Select kernel
-    std::stringstream harris_score_kernel_name;
-    harris_score_kernel_name << "harris_score_" << block_size << "x" << block_size;
-
-    // Create build options
-    std::set<std::string> build_opts = { ("-DDATA_TYPE=" + get_cl_type_from_data_type(input1->info()->data_type())) };
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, harris_score_kernel_name.str(), build_opts);
-
-    // Set static kernel arguments
-    unsigned int idx = 3 * num_arguments_per_2D_tensor(); //Skip the input and output parameters
-    _kernel.setArg(idx++, sensitivity);
-    _kernel.setArg(idx++, strength_thresh);
-    _kernel.setArg(idx++, norm_factor);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 4;
-    constexpr unsigned int num_elems_written_per_iteration   = 4;
-    const unsigned int     num_elems_read_per_iteration      = block_size == 7 ? 10 : 8;
-    const unsigned int     num_rows_read_per_iteration       = block_size;
-
-    Window win = calculate_max_window(*_input1->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input1_access(input1->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowRectangle  input2_access(input2->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input1_access, input2_access, output_access);
-
-    ValidRegion valid_region = intersect_valid_regions(input1->info()->valid_region(), input2->info()->valid_region());
-    output_access.set_valid_region(win, valid_region, border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = harris_score_kernel_name.str();
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input1->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input1->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input1->info()->dimension(1));
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input2->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input2->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input2->info()->dimension(1));
-}
-
-void CLHarrisScoreKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input1, slice);
-        add_2D_tensor_argument(idx, _input2, slice);
-        add_2D_tensor_argument(idx, _output, slice);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLHarrisCornersKernel.h b/src/core/CL/kernels/CLHarrisCornersKernel.h
deleted file mode 100644
index 6482b0aa4e..0000000000
--- a/src/core/CL/kernels/CLHarrisCornersKernel.h
+++ /dev/null
@@ -1,100 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLHARRISCORNERSKERNEL_H
-#define ARM_COMPUTE_CLHARRISCORNERSKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-#include <cstdint>
-
-namespace arm_compute
-{
-class ICLTensor;
-using ICLImage = ICLTensor;
-
-/** Interface for the harris score kernel.
- *
- * @note The implementation supports 3, 5, and 7 for the block_size.
- */
-class CLHarrisScoreKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLHarrisScoreKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHarrisScoreKernel(const CLHarrisScoreKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHarrisScoreKernel &operator=(const CLHarrisScoreKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLHarrisScoreKernel(CLHarrisScoreKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLHarrisScoreKernel &operator=(CLHarrisScoreKernel &&) = default;
-    /** Default destructor */
-    ~CLHarrisScoreKernel() = default;
-
-    /** Setup the kernel parameters
-     *
-     * @param[in]  input1           Source image (gradient X). Data types supported S16, S32. (Must be the same as input2)
-     * @param[in]  input2           Source image (gradient Y). Data types supported S16, S32. (Must be the same as input1)
-     * @param[out] output           Destination image (harris score). Data types supported F32
-     * @param[in]  block_size       The block window size used to compute the Harris Corner score.  Supports: 3, 5 and 7
-     * @param[in]  norm_factor      Normalization factor to use accordingly with the gradient size (Must be different from 0)
-     * @param[in]  strength_thresh  Minimum threshold with which to eliminate Harris Corner scores (computed using the normalized Sobel kernel).
-     * @param[in]  sensitivity      Sensitivity threshold k from the Harris-Stephens equation.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLImage *input1, const ICLImage *input2, ICLImage *output,
-                   int32_t block_size, float norm_factor, float strength_thresh, float sensitivity,
-                   bool border_undefined);
-    /** Setup the kernel parameters
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input1           Source image (gradient X). Data types supported S16, S32. (Must be the same as input2)
-     * @param[in]  input2           Source image (gradient Y). Data types supported S16, S32. (Must be the same as input1)
-     * @param[out] output           Destination image (harris score). Data types supported F32
-     * @param[in]  block_size       The block window size used to compute the Harris Corner score.  Supports: 3, 5 and 7
-     * @param[in]  norm_factor      Normalization factor to use accordingly with the gradient size (Must be different from 0)
-     * @param[in]  strength_thresh  Minimum threshold with which to eliminate Harris Corner scores (computed using the normalized Sobel kernel).
-     * @param[in]  sensitivity      Sensitivity threshold k from the Harris-Stephens equation.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLImage *input1, const ICLImage *input2, ICLImage *output,
-                   int32_t block_size, float norm_factor, float strength_thresh, float sensitivity,
-                   bool border_undefined);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-    BorderSize border_size() const override;
-
-protected:
-    const ICLImage *_input1;          /**< Source image - Gx component */
-    const ICLImage *_input2;          /**< Source image - Gy component */
-    ICLImage       *_output;          /**< Source image - Harris score */
-    float           _sensitivity;     /**< Sensitivity value */
-    float           _strength_thresh; /**< Threshold value */
-    float           _norm_factor;     /**< Normalization factor */
-    BorderSize      _border_size;     /**< Border size */
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLHARRISCORNERSKERNEL_H */
diff --git a/src/core/CL/kernels/CLHistogramKernel.cpp b/src/core/CL/kernels/CLHistogramKernel.cpp
deleted file mode 100644
index ca5322aa51..0000000000
--- a/src/core/CL/kernels/CLHistogramKernel.cpp
+++ /dev/null
@@ -1,253 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLHistogramKernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLDistribution1D.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <cstring>
-#include <string>
-
-using namespace arm_compute;
-
-// each thread handle 16 pixels
-constexpr signed int pixels_per_item = 16;
-
-// local work group size in X dimension
-constexpr unsigned int local_x_size = 16;
-
-CLHistogramKernel::CLHistogramKernel()
-    : _input(nullptr), _output(nullptr)
-{
-}
-
-void CLHistogramKernel::configure(const ICLImage *input, ICLDistribution1D *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output);
-}
-
-void CLHistogramKernel::configure(const CLCompileContext &compile_context, const ICLImage *input, ICLDistribution1D *output)
-{
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(input);
-    ARM_COMPUTE_ERROR_ON(nullptr == output);
-
-    // Check input size
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-
-    // Check offset
-    ARM_COMPUTE_ERROR_ON_MSG(0 > output->offset() || output->offset() > 256, "Offset is larger than the image value range.");
-
-    // Check range
-    ARM_COMPUTE_ERROR_ON_MSG(output->range() > 256 /* max range */, "Range larger than the image value range.");
-
-    _input  = input;
-    _output = output;
-
-    if(_input->info()->dimension(0) < pixels_per_item)
-    {
-        return;
-    }
-
-    unsigned int num_bins    = _output->num_bins();
-    unsigned int window_size = _output->window();
-    unsigned int offset      = _output->offset();
-    unsigned int range       = _output->range();
-    unsigned int offrange    = offset + range;
-    unsigned int bin_size    = _output->size();
-    unsigned int buffer_size = bin_size + 1; // We need one extra place for pixels that don't meet the conditions
-
-    // Create kernel
-    bool              is_fixed_size = (256 == num_bins) && (1 == window_size) && (0 == offset) && (256 == offrange);
-    const std::string kernel_name   = is_fixed_size ? "hist_local_kernel_fixed" : "hist_local_kernel";
-    _kernel                         = create_kernel(compile_context, kernel_name);
-
-    // Set static kernel arguments
-    unsigned int idx = num_arguments_per_2D_tensor(); //Skip the input and output parameters
-    _kernel.setArg(idx++, buffer_size, nullptr);
-    _kernel.setArg(idx++, _output->cl_buffer());
-    if(!is_fixed_size)
-    {
-        _kernel.setArg<cl_uint>(idx++, num_bins);
-        _kernel.setArg<cl_uint>(idx++, offset);
-        _kernel.setArg<cl_uint>(idx++, range);
-        _kernel.setArg<cl_uint>(idx++, offrange);
-    }
-
-    // We only run histogram on Image, therefore only 2 dimensions here
-    unsigned int end_position = (_input->info()->dimension(0) / pixels_per_item) * pixels_per_item;
-
-    // Configure kernel window
-    Window win;
-    win.set(0, Window::Dimension(0, end_position, pixels_per_item));
-    win.set(1, Window::Dimension(0, _input->info()->dimension(1)));
-
-    update_window_and_padding(win, AccessWindowHorizontal(input->info(), 0, pixels_per_item));
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-}
-
-void CLHistogramKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(ICLKernel::window(), window);
-
-    // TODO (COMPMID-679): Add CLMemFill
-    _output->map(queue, true);
-    ARM_COMPUTE_ERROR_ON(_output->buffer() == nullptr);
-    memset(_output->buffer(), 0, _output->size());
-    _output->unmap(queue);
-
-    if(_input->info()->dimension(0) < pixels_per_item)
-    {
-        return;
-    }
-
-    Window             slice = window.first_slice_window_2D();
-    const unsigned int gws_x = (window.x().end() - window.x().start()) / window.x().step();
-    cl::NDRange        lws   = (local_x_size < gws_x) ? cl::NDRange(local_x_size, 1) : cl::NDRange(1, 1);
-
-    do
-    {
-        /* Run the core part which has width can be divided by 16 */
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-
-        enqueue(queue, *this, slice, lws);
-    }
-    while(window.slide_window_slice_2D(slice));
-}
-
-CLHistogramBorderKernel::CLHistogramBorderKernel()
-    : _input(nullptr), _output(nullptr)
-{
-}
-
-void CLHistogramBorderKernel::configure(const ICLImage *input, ICLDistribution1D *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output);
-}
-
-void CLHistogramBorderKernel::configure(const CLCompileContext &compile_context, const ICLImage *input, ICLDistribution1D *output)
-{
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(input);
-    ARM_COMPUTE_ERROR_ON(nullptr == output);
-
-    // Check input size
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-
-    // Check offset
-    ARM_COMPUTE_ERROR_ON_MSG(0 > output->offset() || output->offset() > 256, "Offset is larger than the image value range.");
-
-    // Check range
-    ARM_COMPUTE_ERROR_ON_MSG(output->range() > 256 /* max range */, "Range larger than the image value range.");
-
-    // We only run histogram on Image, therefore only 2 dimensions here
-    unsigned int start_position = (input->info()->dimension(0) / pixels_per_item) * pixels_per_item;
-
-    if(start_position >= input->info()->dimension(0))
-    {
-        return; // no need to run histogram border kernel
-    }
-
-    _input  = input;
-    _output = output;
-
-    unsigned int num_bins    = _output->num_bins();
-    unsigned int window_size = _output->window();
-    unsigned int offset      = _output->offset();
-    unsigned int range       = _output->range();
-    unsigned int offrange    = offset + range;
-
-    // Create kernel
-    bool              is_fixed_size = (256 == num_bins) && (1 == window_size) && (0 == offset) && (256 == offrange);
-    const std::string kernel_name   = is_fixed_size ? "hist_border_kernel_fixed" : "hist_border_kernel";
-    _kernel                         = create_kernel(compile_context, kernel_name);
-
-    // Set static kernel arguments
-    unsigned int idx = num_arguments_per_2D_tensor(); //Skip the input and output parameters
-    _kernel.setArg(idx++, _output->cl_buffer());
-    if(!is_fixed_size)
-    {
-        _kernel.setArg<cl_uint>(idx++, num_bins);
-        _kernel.setArg<cl_uint>(idx++, offset);
-        _kernel.setArg<cl_uint>(idx++, range);
-        _kernel.setArg<cl_uint>(idx++, offrange);
-    }
-
-    // Configure kernel window
-    Window win;
-    win.set(0, Window::Dimension(start_position, _input->info()->dimension(0)));
-    win.set(1, Window::Dimension(0, _input->info()->dimension(1)));
-    update_window_and_padding(win, AccessWindowHorizontal(input->info(), 0, 1));
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-}
-
-void CLHistogramBorderKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    if(window.x().start() >= window.x().end())
-    {
-        return;
-    }
-
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(ICLKernel::window(), window);
-
-    cl::NDRange lws = cl::NDRange(1, 1);
-
-    Window slice = window.first_slice_window_2D();
-
-    do
-    {
-        /* Run the border part which has width cannot be divided by 16 */
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-
-        enqueue(queue, *this, slice, lws);
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLHistogramKernel.h b/src/core/CL/kernels/CLHistogramKernel.h
deleted file mode 100644
index 9c97c6590d..0000000000
--- a/src/core/CL/kernels/CLHistogramKernel.h
+++ /dev/null
@@ -1,111 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLHISTOGRAMKERNEL_H
-#define ARM_COMPUTE_CLHISTOGRAMKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLDistribution1D;
-class ICLTensor;
-using ICLImage = ICLTensor;
-
-/** Interface to run the histogram kernel. This kernel processes the part of image with width can be divided by 16.
- *  If the image width is not a multiple of 16, remaining pixels have to be processed with the @ref CLHistogramBorderKernel
- */
-class CLHistogramKernel : public ICLKernel
-{
-public:
-    /** Constructor */
-    CLHistogramKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHistogramKernel(const CLHistogramKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHistogramKernel &operator=(const CLHistogramKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLHistogramKernel(CLHistogramKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLHistogramKernel &operator=(CLHistogramKernel &&) = default;
-    /** Initialise the kernel's input, output and border mode.
-     *
-     * @param[in]  input  Source image. Data types supported: U8.
-     * @param[out] output Destination distribution.
-     */
-    void configure(const ICLImage *input, ICLDistribution1D *output);
-    /** Initialise the kernel's input, output and border mode.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Source image. Data types supported: U8.
-     * @param[out] output          Destination distribution.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLImage *input, ICLDistribution1D *output);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLImage    *_input;
-    ICLDistribution1D *_output;
-};
-
-/** Interface to run the histogram kernel to handle the leftover part of image
- *
- */
-class CLHistogramBorderKernel : public ICLKernel
-{
-public:
-    /** Constructor */
-    CLHistogramBorderKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHistogramBorderKernel(const CLHistogramBorderKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHistogramBorderKernel &operator=(const CLHistogramBorderKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLHistogramBorderKernel(CLHistogramBorderKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLHistogramBorderKernel &operator=(CLHistogramBorderKernel &&) = default;
-    /** Initialise the kernel's input, output and border mode.
-     *
-     * @param[in]  input  Source image. Data types supported: U8.
-     * @param[out] output Destination distribution.
-     */
-    void configure(const ICLImage *input, ICLDistribution1D *output);
-    /** Initialise the kernel's input, output and border mode.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Source image. Data types supported: U8.
-     * @param[out] output          Destination distribution.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLImage *input, ICLDistribution1D *output);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLImage    *_input;
-    ICLDistribution1D *_output;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLHISTOGRAMKERNEL_H*/
diff --git a/src/core/CL/kernels/CLIntegralImageKernel.cpp b/src/core/CL/kernels/CLIntegralImageKernel.cpp
deleted file mode 100644
index 5e5683d231..0000000000
--- a/src/core/CL/kernels/CLIntegralImageKernel.cpp
+++ /dev/null
@@ -1,146 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLIntegralImageKernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <cstddef>
-
-using namespace arm_compute;
-
-void CLIntegralImageHorKernel::configure(const ICLTensor *input, ICLTensor *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output);
-}
-
-void CLIntegralImageHorKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U32);
-
-    _input  = input;
-    _output = output;
-
-    // Create kernel
-    const std::string kernel_name = std::string("integral_horizontal");
-    _kernel                       = create_kernel(compile_context, kernel_name);
-
-    // Configure kernel window
-    const unsigned int num_elems_processed_per_iteration = input->info()->dimension(0);
-    const unsigned int num_elems_accessed_per_iteration  = ceil_to_multiple(num_elems_processed_per_iteration, 16);
-
-    Window                 win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration));
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_accessed_per_iteration);
-
-    update_window_and_padding(win,
-                              AccessWindowHorizontal(input->info(), 0, num_elems_accessed_per_iteration),
-                              output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-}
-
-CLIntegralImageVertKernel::CLIntegralImageVertKernel()
-    : _in_out(nullptr)
-{
-}
-
-void CLIntegralImageVertKernel::configure(ICLTensor *in_out)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), in_out);
-}
-
-void CLIntegralImageVertKernel::configure(const CLCompileContext &compile_context, ICLTensor *in_out)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(in_out, 1, DataType::U32);
-
-    _in_out = in_out;
-
-    // Create kernel
-    const std::string kernel_name = std::string("integral_vertical");
-    _kernel                       = create_kernel(compile_context, kernel_name);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration_x = 8;
-    const unsigned int     num_elems_processed_per_iteration_y = in_out->info()->dimension(Window::DimY);
-
-    Window win = calculate_max_window(*in_out->info(), Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-
-    AccessWindowRectangle in_out_access(in_out->info(), 0, 0, num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y);
-
-    update_window_and_padding(win, in_out_access);
-
-    in_out_access.set_valid_region(win, in_out->info()->valid_region());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(in_out->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(in_out->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(in_out->info()->dimension(1));
-}
-
-void CLIntegralImageVertKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    const size_t height = _in_out->info()->dimension(1);
-
-    Window slice = window.first_slice_window_2D();
-
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _in_out, slice);
-        _kernel.setArg<cl_uint>(idx++, height);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLIntegralImageKernel.h b/src/core/CL/kernels/CLIntegralImageKernel.h
deleted file mode 100644
index 0e40e3afbc..0000000000
--- a/src/core/CL/kernels/CLIntegralImageKernel.h
+++ /dev/null
@@ -1,86 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLINTEGRALIMAGEKERNEL_H
-#define ARM_COMPUTE_CLINTEGRALIMAGEKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface to run the horizontal pass of the integral image kernel. */
-class CLIntegralImageHorKernel : public ICLSimple2DKernel
-{
-public:
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in]  input  An input tensor. Data types supported: U8
-     * @param[out] output Destination tensor, Data types supported: U32.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output);
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           An input tensor. Data types supported: U8
-     * @param[out] output          Destination tensor, Data types supported: U32.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output);
-};
-
-/** Interface to run the vertical pass of the integral image kernel. */
-class CLIntegralImageVertKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLIntegralImageVertKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLIntegralImageVertKernel(const CLIntegralImageVertKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLIntegralImageVertKernel &operator=(const CLIntegralImageVertKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLIntegralImageVertKernel(CLIntegralImageVertKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLIntegralImageVertKernel &operator=(CLIntegralImageVertKernel &&) = default;
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in,out] in_out The input/output tensor. Data types supported: U32
-     */
-    void configure(ICLTensor *in_out);
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in]     compile_context The compile context to be used.
-     * @param[in,out] in_out          The input/output tensor. Data types supported: U32
-     */
-    void configure(const CLCompileContext &compile_context, ICLTensor *in_out);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    ICLTensor *_in_out;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLINTEGRALIMAGEKERNEL_H */
diff --git a/src/core/CL/kernels/CLMagnitudePhaseKernel.cpp b/src/core/CL/kernels/CLMagnitudePhaseKernel.cpp
deleted file mode 100644
index 9845dd6169..0000000000
--- a/src/core/CL/kernels/CLMagnitudePhaseKernel.cpp
+++ /dev/null
@@ -1,176 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLMagnitudePhaseKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <set>
-#include <string>
-
-using namespace arm_compute;
-
-CLMagnitudePhaseKernel::CLMagnitudePhaseKernel()
-    : _gx(nullptr), _gy(nullptr), _magnitude(nullptr), _phase(nullptr), _run_mag(false), _run_phase(false)
-{
-}
-
-void CLMagnitudePhaseKernel::configure(const ICLTensor *gx, const ICLTensor *gy, ICLTensor *magnitude, ICLTensor *phase,
-                                       MagnitudeType mag_type, PhaseType phase_type)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), gx, gy, magnitude, phase, mag_type, phase_type);
-}
-
-void CLMagnitudePhaseKernel::configure(const CLCompileContext &compile_context, const ICLTensor *gx, const ICLTensor *gy, ICLTensor *magnitude, ICLTensor *phase,
-                                       MagnitudeType mag_type, PhaseType phase_type)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(gx, 1, DataType::S16, DataType::S32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(gy, 1, DataType::S16, DataType::S32);
-    ARM_COMPUTE_ERROR_ON((magnitude == nullptr) && (phase == nullptr));
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(gx, gy);
-
-    _run_mag   = (magnitude != nullptr);
-    _run_phase = (phase != nullptr);
-    if(_run_mag)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(magnitude, 1, DataType::S16, DataType::S32);
-        ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(gx, magnitude);
-    }
-    if(_run_phase)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(phase, 1, DataType::U8);
-    }
-
-    if(!_run_mag && !_run_phase)
-    {
-        ARM_COMPUTE_ERROR("At least one output must be NOT NULL");
-    }
-
-    _gx        = gx;
-    _gy        = gy;
-    _magnitude = magnitude;
-    _phase     = phase;
-
-    // Construct kernel name
-    std::set<std::string> build_opts = {};
-
-    // Add magnitude type
-    if(_run_mag)
-    {
-        switch(mag_type)
-        {
-            case MagnitudeType::L1NORM:
-                build_opts.insert("-DMAGNITUDE=1");
-                break;
-            case MagnitudeType::L2NORM:
-                build_opts.insert("-DMAGNITUDE=2");
-                break;
-            default:
-                ARM_COMPUTE_ERROR("Unsupported magnitude calculation type.");
-                build_opts.insert("-DMAGNITUDE=0");
-                break;
-        }
-    }
-
-    // Add phase type
-    if(_run_phase)
-    {
-        switch(phase_type)
-        {
-            case PhaseType::UNSIGNED:
-                build_opts.insert("-DPHASE=1");
-                break;
-            case PhaseType::SIGNED:
-                build_opts.insert("-DPHASE=2");
-                break;
-            default:
-                ARM_COMPUTE_ERROR("Unsupported phase calculation type.");
-                build_opts.insert("-DPHASE=0");
-                break;
-        }
-    }
-
-    // Add data_type
-    build_opts.insert("-DDATA_TYPE=" + get_cl_type_from_data_type(gx->info()->data_type()));
-
-    // Create kernel
-    const std::string kernel_name = std::string("magnitude_phase");
-    _kernel                       = create_kernel(compile_context, kernel_name, build_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 16;
-
-    Window win = calculate_max_window(*gx->info(), Steps(num_elems_processed_per_iteration));
-
-    AccessWindowHorizontal gx_access(gx->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal gy_access(gy->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal output_magnitude_access(magnitude == nullptr ? nullptr : magnitude->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal output_phase_access(phase == nullptr ? nullptr : phase->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win,
-                              gx_access, gy_access,
-                              output_magnitude_access, output_phase_access);
-
-    ValidRegion valid_region = intersect_valid_regions(gx->info()->valid_region(),
-                                                       gy->info()->valid_region());
-    output_magnitude_access.set_valid_region(win, valid_region);
-    output_phase_access.set_valid_region(win, valid_region);
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(gx->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(gx->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(gx->info()->dimension(1));
-}
-
-void CLMagnitudePhaseKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _gx, slice);
-        add_2D_tensor_argument(idx, _gy, slice);
-        add_2D_tensor_argument_if((_run_mag), idx, _magnitude, slice);
-        add_2D_tensor_argument_if((_run_phase), idx, _phase, slice);
-
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLMagnitudePhaseKernel.h b/src/core/CL/kernels/CLMagnitudePhaseKernel.h
deleted file mode 100644
index 514036b2ff..0000000000
--- a/src/core/CL/kernels/CLMagnitudePhaseKernel.h
+++ /dev/null
@@ -1,90 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLMAGNITUDEPHASEKERNEL_H
-#define ARM_COMPUTE_CLMAGNITUDEPHASEKERNEL_H
-
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Template interface for the kernel to compute magnitude and phase.
- *
- */
-class CLMagnitudePhaseKernel : public ICLKernel
-{
-public:
-    /** Default constructor. */
-    CLMagnitudePhaseKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLMagnitudePhaseKernel(const CLMagnitudePhaseKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLMagnitudePhaseKernel &operator=(const CLMagnitudePhaseKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLMagnitudePhaseKernel(CLMagnitudePhaseKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLMagnitudePhaseKernel &operator=(CLMagnitudePhaseKernel &&) = default;
-    /** Initialise the kernel's input, output.
-     *
-     * @note At least one of output1 or output2 must be set.
-     *
-     * @param[in]  gx         The input gradient X tensor. Data types supported: S16/S32.
-     * @param[in]  gy         The input gradient Y tensor. Data types supported: S16/S32.
-     * @param[out] magnitude  (Optional) The output tensor - Magnitude. Data types supported: S16/S32.
-     * @param[out] phase      (Optional) The output tensor - Phase. Data types supported: U8.
-     * @param[in]  mag_type   (Optional) Magnitude calculation type. Default: L2NORM.
-     * @param[in]  phase_type (Optional) Phase calculation type. Default: SIGNED.
-     */
-    void configure(const ICLTensor *gx, const ICLTensor *gy, ICLTensor *magnitude, ICLTensor *phase,
-                   MagnitudeType mag_type = MagnitudeType::L2NORM, PhaseType phase_type = PhaseType::SIGNED);
-    /** Initialise the kernel's input, output.
-     *
-     * @note At least one of output1 or output2 must be set.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  gx              The input gradient X tensor. Data types supported: S16/S32.
-     * @param[in]  gy              The input gradient Y tensor. Data types supported: S16/S32.
-     * @param[out] magnitude       (Optional) The output tensor - Magnitude. Data types supported: S16/S32.
-     * @param[out] phase           (Optional) The output tensor - Phase. Data types supported: U8.
-     * @param[in]  mag_type        (Optional) Magnitude calculation type. Default: L2NORM.
-     * @param[in]  phase_type      (Optional) Phase calculation type. Default: SIGNED.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *gx, const ICLTensor *gy, ICLTensor *magnitude, ICLTensor *phase,
-                   MagnitudeType mag_type = MagnitudeType::L2NORM, PhaseType phase_type = PhaseType::SIGNED);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_gx;        /**< Input gradient X. */
-    const ICLTensor *_gy;        /**< Input gradient Y. */
-    ICLTensor       *_magnitude; /**< Output - Magnitude. */
-    ICLTensor       *_phase;     /**< Output - Phase. */
-    bool             _run_mag;   /**< Calculate magnitude ? */
-    bool             _run_phase; /**< Calculate phase ? */
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLMAGNITUDEPHASEKERNEL_H */
diff --git a/src/core/CL/kernels/CLMeanStdDevKernel.cpp b/src/core/CL/kernels/CLMeanStdDevKernel.cpp
deleted file mode 100644
index aed6e6eaf7..0000000000
--- a/src/core/CL/kernels/CLMeanStdDevKernel.cpp
+++ /dev/null
@@ -1,156 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLMeanStdDevKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "src/core/CL/CLValidate.h"
-#include "src/core/helpers/WindowHelpers.h"
-
-#include <cmath>
-#include <set>
-#include <string>
-
-using namespace arm_compute;
-
-CLMeanStdDevKernel::CLMeanStdDevKernel()
-    : _input(nullptr), _mean(nullptr), _stddev(nullptr), _global_sum(nullptr), _global_sum_squared(nullptr), _border_size(0)
-{
-}
-
-BorderSize CLMeanStdDevKernel::border_size() const
-{
-    return _border_size;
-}
-
-Status CLMeanStdDevKernel::validate(const ITensorInfo *input, float *mean, cl::Buffer *global_sum, float *stddev, cl::Buffer *global_sum_squared)
-{
-    ARM_COMPUTE_UNUSED(mean);
-    ARM_COMPUTE_UNUSED(stddev);
-    ARM_COMPUTE_UNUSED(global_sum);
-    ARM_COMPUTE_UNUSED(global_sum_squared);
-    ARM_COMPUTE_RETURN_ERROR_ON_INT64_BASE_ATOMICS_UNSUPPORTED();
-    ARM_COMPUTE_RETURN_ERROR_ON_TENSOR_NOT_2D(input);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-
-    return Status{};
-}
-
-void CLMeanStdDevKernel::configure(const ICLImage *input, float *mean, cl::Buffer *global_sum, float *stddev, cl::Buffer *global_sum_squared)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, mean, global_sum, stddev, global_sum_squared);
-}
-
-void CLMeanStdDevKernel::configure(const CLCompileContext &compile_context, const ICLImage *input, float *mean, cl::Buffer *global_sum, float *stddev, cl::Buffer *global_sum_squared)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input, mean, global_sum);
-    ARM_COMPUTE_ERROR_ON(stddev && nullptr == global_sum_squared);
-    ARM_COMPUTE_ERROR_THROW_ON(CLMeanStdDevKernel::validate(input->info(), mean, global_sum, stddev, global_sum_squared));
-
-    _input              = input;
-    _mean               = mean;
-    _stddev             = stddev;
-    _global_sum         = global_sum;
-    _global_sum_squared = global_sum_squared;
-
-    // Create kernel
-    std::set<std::string> build_opts;
-
-    if(_stddev != nullptr)
-    {
-        build_opts.insert("-DSTDDEV");
-    }
-
-    _kernel = create_kernel(compile_context, "mean_stddev_accumulate", build_opts);
-
-    // Set fixed arguments
-    unsigned int idx = num_arguments_per_2D_tensor(); //Skip the input parameters
-
-    _kernel.setArg(idx++, static_cast<cl_uint>(input->info()->dimension(1)));
-    _kernel.setArg(idx++, *_global_sum);
-
-    if(_stddev != nullptr)
-    {
-        _kernel.setArg(idx++, *_global_sum_squared);
-    }
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration_x = 8;
-    const unsigned int     num_elems_processed_per_iteration_y = input->info()->dimension(1);
-
-    _border_size = BorderSize(ceil_to_multiple(input->info()->dimension(0), num_elems_processed_per_iteration_x) - input->info()->dimension(0));
-
-    Window                win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-    AccessWindowRectangle input_access(input->info(), 0, 0, num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y);
-    update_window_and_padding(win, input_access);
-
-    ICLKernel::configure_internal(win);
-}
-
-void CLMeanStdDevKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    // Clear sums
-    static const cl_ulong zero = 0;
-    queue.enqueueWriteBuffer(*_global_sum, CL_FALSE, 0, sizeof(cl_ulong), &zero);
-
-    if(_stddev != nullptr)
-    {
-        queue.enqueueWriteBuffer(*_global_sum_squared, CL_FALSE, 0, sizeof(cl_ulong), &zero);
-    }
-
-    Window slice = window.first_slice_window_2D();
-
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-        // Set slice step equal to height to force gws[1] to 1,
-        // as each thread calculates the sum across all rows and columns equal to the number of elements processed by each work-item
-        slice.set_dimension_step(Window::DimY, _input->info()->dimension(1));
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-
-    // Calculate mean and stddev
-    cl_ulong    global_sum         = 0;
-    cl_ulong    global_sum_squared = 0;
-    const float num_pixels         = _input->info()->dimension(0) * _input->info()->dimension(1);
-
-    queue.enqueueReadBuffer(*_global_sum, CL_TRUE, 0, sizeof(cl_ulong), static_cast<void *>(&global_sum));
-    const float mean = global_sum / num_pixels;
-    *_mean           = mean;
-
-    if(_stddev != nullptr)
-    {
-        queue.enqueueReadBuffer(*_global_sum_squared, CL_TRUE, 0, sizeof(cl_ulong), static_cast<void *>(&global_sum_squared));
-        *_stddev = std::sqrt((global_sum_squared / num_pixels) - (mean * mean));
-    }
-}
diff --git a/src/core/CL/kernels/CLMeanStdDevKernel.h b/src/core/CL/kernels/CLMeanStdDevKernel.h
deleted file mode 100644
index 179a2025b7..0000000000
--- a/src/core/CL/kernels/CLMeanStdDevKernel.h
+++ /dev/null
@@ -1,98 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLMEANSTDDEVKERNEL_H
-#define ARM_COMPUTE_CLMEANSTDDEVKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace cl
-{
-class Buffer;
-}
-
-namespace arm_compute
-{
-class ICLTensor;
-using ICLImage = ICLTensor;
-
-/** Interface for the kernel to calculate mean and standard deviation of input image pixels. */
-class CLMeanStdDevKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLMeanStdDevKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLMeanStdDevKernel(const CLMeanStdDevKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLMeanStdDevKernel &operator=(const CLMeanStdDevKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLMeanStdDevKernel(CLMeanStdDevKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLMeanStdDevKernel &operator=(CLMeanStdDevKernel &&) = default;
-    /** Initialise the kernel's input and outputs.
-     *
-     * @param[in]  input              Input image. Data types supported: U8.
-     * @param[out] mean               Input average pixel value.
-     * @param[out] global_sum         Keeps global sum of pixel values (Buffer size: 1 cl_ulong).
-     * @param[out] stddev             (Optional) Output standard deviation of pixel values.
-     * @param[out] global_sum_squared (Optional if stddev is not set, required if stddev is set) Keeps global sum of squared pixel values (Buffer size: 1 cl_ulong).
-     */
-    void configure(const ICLImage *input, float *mean, cl::Buffer *global_sum, float *stddev = nullptr, cl::Buffer *global_sum_squared = nullptr);
-    /** Initialise the kernel's input and outputs.
-     *
-     * @param[in]  compile_context    The compile context to be used.
-     * @param[in]  input              Input image. Data types supported: U8.
-     * @param[out] mean               Input average pixel value.
-     * @param[out] global_sum         Keeps global sum of pixel values (Buffer size: 1 cl_ulong).
-     * @param[out] stddev             (Optional) Output standard deviation of pixel values.
-     * @param[out] global_sum_squared (Optional if stddev is not set, required if stddev is set) Keeps global sum of squared pixel values (Buffer size: 1 cl_ulong).
-     */
-    void configure(const CLCompileContext &compile_context, const ICLImage *input, float *mean, cl::Buffer *global_sum, float *stddev = nullptr, cl::Buffer *global_sum_squared = nullptr);
-    /** Static function to check if given info will lead to a valid configuration of @ref CLMeanStdDevKernel.
-     *
-     * @param[in] input              Input image info. Data types supported: U8.
-     * @param[in] mean               Input average pixel value.
-     * @param[in] global_sum         Keeps global sum of pixel values.
-     * @param[in] stddev             (Optional) Output standard deviation of pixel values.
-     * @param[in] global_sum_squared (Optional if stddev is not set, required if stddev is set) Keeps global sum of squared pixel values.
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input, float *mean, cl::Buffer *global_sum, float *stddev = nullptr, cl::Buffer *global_sum_squared = nullptr);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-    BorderSize border_size() const override;
-
-private:
-    const ICLImage *_input;
-    float          *_mean;
-    float          *_stddev;
-    cl::Buffer     *_global_sum;
-    cl::Buffer     *_global_sum_squared;
-    BorderSize      _border_size;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLMEANSTDDEVKERNEL_H */
diff --git a/src/core/CL/kernels/CLMedian3x3Kernel.cpp b/src/core/CL/kernels/CLMedian3x3Kernel.cpp
deleted file mode 100644
index 23a21d6b19..0000000000
--- a/src/core/CL/kernels/CLMedian3x3Kernel.cpp
+++ /dev/null
@@ -1,88 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLMedian3x3Kernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-using namespace arm_compute;
-
-BorderSize CLMedian3x3Kernel::border_size() const
-{
-    return BorderSize(1);
-}
-
-void CLMedian3x3Kernel::configure(const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, border_undefined);
-}
-
-void CLMedian3x3Kernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8);
-
-    _input  = input;
-    _output = output;
-
-    // Create kernel
-    const std::string kernel_name = std::string("non_linear_filter_box3x3");
-    _kernel                       = create_kernel(compile_context, kernel_name, { "-DMEDIAN" });
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-    constexpr unsigned int num_rows_read_per_iteration       = 3;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(border_undefined);
-}
diff --git a/src/core/CL/kernels/CLMedian3x3Kernel.h b/src/core/CL/kernels/CLMedian3x3Kernel.h
deleted file mode 100644
index 8cc5ed7279..0000000000
--- a/src/core/CL/kernels/CLMedian3x3Kernel.h
+++ /dev/null
@@ -1,59 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLMEDIAN3X3KERNEL_H
-#define ARM_COMPUTE_CLMEDIAN3X3KERNEL_H
-
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the median 3x3 filter kernel.
- *
- */
-class CLMedian3x3Kernel : public ICLSimple2DKernel
-{
-public:
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in]  input            An input tensor. Data types supported: U8
-     * @param[out] output           The output tensor. Data types supported: U8.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, bool border_undefined);
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            An input tensor. Data types supported: U8
-     * @param[out] output           The output tensor. Data types supported: U8.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined);
-
-    // Inherited methods overridden:
-    BorderSize border_size() const override;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLMEDIAN3X3KERNEL_H */
diff --git a/src/core/CL/kernels/CLMinMaxLocationKernel.cpp b/src/core/CL/kernels/CLMinMaxLocationKernel.cpp
deleted file mode 100644
index 675cfc19a9..0000000000
--- a/src/core/CL/kernels/CLMinMaxLocationKernel.cpp
+++ /dev/null
@@ -1,246 +0,0 @@
-/*
- * Copyright (c) 2017-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLMinMaxLocationKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/AutoConfiguration.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <climits>
-
-namespace arm_compute
-{
-inline int32_t FloatFlip(float val)
-{
-    static_assert(sizeof(float) == sizeof(int32_t), "Float must be same size as int32_t");
-    int32_t int_val = 0;
-
-    memcpy(&int_val, &val, sizeof(float));
-    int_val = (int_val >= 0) ? int_val : int_val ^ 0x7FFFFFFF;
-    return int_val;
-}
-
-inline float IFloatFlip(int32_t val)
-{
-    static_assert(sizeof(float) == sizeof(int32_t), "Float must be same size as int32_t");
-    float flt_val = 0.f;
-
-    val = (val >= 0) ? val : val ^ 0x7FFFFFFF;
-    memcpy(&flt_val, &val, sizeof(float));
-    return flt_val;
-}
-
-CLMinMaxKernel::CLMinMaxKernel()
-    : _input(nullptr), _min_max(), _data_type_max_min()
-{
-}
-
-void CLMinMaxKernel::configure(const ICLImage *input, cl::Buffer *min_max)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, min_max);
-}
-
-void CLMinMaxKernel::configure(const CLCompileContext &compile_context, const ICLImage *input, cl::Buffer *min_max)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S16, DataType::F32);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(input);
-    ARM_COMPUTE_ERROR_ON(min_max == nullptr);
-
-    _input                                               = input;
-    _min_max                                             = min_max;
-    const unsigned int num_elems_processed_per_iteration = input->info()->dimension(0);
-
-    switch(input->info()->data_type())
-    {
-        case DataType::U8:
-            _data_type_max_min[0] = UCHAR_MAX;
-            _data_type_max_min[1] = 0;
-            break;
-        case DataType::S16:
-            _data_type_max_min[0] = SHRT_MAX;
-            _data_type_max_min[1] = SHRT_MIN;
-            break;
-        case DataType::F32:
-            _data_type_max_min[0] = FloatFlip(std::numeric_limits<float>::max());
-            _data_type_max_min[1] = FloatFlip(std::numeric_limits<float>::lowest());
-            break;
-        default:
-            ARM_COMPUTE_ERROR("You called with the wrong image data types");
-    }
-
-    // Set kernel build options
-    std::set<std::string> build_opts{ "-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()) };
-
-    if(num_elems_processed_per_iteration % max_cl_vector_width != 0)
-    {
-        build_opts.emplace("-DNON_MULTIPLE_OF_16");
-    }
-
-    if(input->info()->data_type() == DataType::F32)
-    {
-        build_opts.emplace("-DDATA_TYPE_MAX=" + support::cpp11::to_string(std::numeric_limits<float>::max()));
-        build_opts.emplace("-DDATA_TYPE_MIN=" + support::cpp11::to_string(std::numeric_limits<float>::lowest()));
-        build_opts.emplace("-DIS_DATA_TYPE_FLOAT");
-    }
-    else
-    {
-        build_opts.emplace("-DDATA_TYPE_MAX=" + support::cpp11::to_string(_data_type_max_min[0]));
-        build_opts.emplace("-DDATA_TYPE_MIN=" + support::cpp11::to_string(_data_type_max_min[1]));
-    }
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "minmax", build_opts);
-
-    // Set fixed arguments
-    unsigned int idx = num_arguments_per_2D_tensor(); //Skip the input and output parameters
-    _kernel.setArg(idx++, *_min_max);
-    _kernel.setArg<cl_int>(idx++, static_cast<cl_int>(input->info()->dimension(0)));
-
-    // Configure kernel window
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration));
-    update_window_and_padding(win, AccessWindowHorizontal(input->info(), 0, ceil_to_multiple(num_elems_processed_per_iteration, 16)));
-    ICLKernel::configure_internal(win);
-}
-
-void CLMinMaxKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
-
-    // Reset mininum and maximum values
-    queue.enqueueWriteBuffer(*_min_max, CL_FALSE /* blocking */, 0, _data_type_max_min.size() * sizeof(int), _data_type_max_min.data());
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-
-    cl_int min = 0;
-    cl_int max = 0;
-    queue.enqueueReadBuffer(*_min_max, CL_TRUE /* blocking */, 0 * sizeof(cl_int), sizeof(cl_int), static_cast<int *>(&min));
-    queue.enqueueReadBuffer(*_min_max, CL_TRUE /* blocking */, 1 * sizeof(cl_int), sizeof(cl_int), static_cast<int *>(&max));
-
-    if(_input->info()->data_type() == DataType::F32)
-    {
-        std::array<float, 2> min_max =
-        {
-            {
-                IFloatFlip(min),
-                IFloatFlip(max)
-            }
-        };
-        queue.enqueueWriteBuffer(*_min_max, CL_TRUE /* blocking */, 0, min_max.size() * sizeof(float), min_max.data());
-    }
-    else
-    {
-        std::array<int32_t, 2> min_max = { { min, max } };
-        queue.enqueueWriteBuffer(*_min_max, CL_TRUE /* blocking */, 0, min_max.size() * sizeof(int32_t), min_max.data());
-    }
-}
-
-CLMinMaxLocationKernel::CLMinMaxLocationKernel()
-    : _input(nullptr), _min_max_count(nullptr)
-{
-}
-
-void CLMinMaxLocationKernel::configure(const ICLImage *input, cl::Buffer *min_max, cl::Buffer *min_max_count, ICLCoordinates2DArray *min_loc, ICLCoordinates2DArray *max_loc)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, min_max, min_max_count, min_loc, max_loc);
-}
-
-void CLMinMaxLocationKernel::configure(const CLCompileContext &compile_context, const ICLImage *input, cl::Buffer *min_max, cl::Buffer *min_max_count, ICLCoordinates2DArray *min_loc,
-                                       ICLCoordinates2DArray *max_loc)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S16, DataType::F32);
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(input);
-    ARM_COMPUTE_ERROR_ON(min_max == nullptr);
-    ARM_COMPUTE_ERROR_ON(min_max_count == nullptr && min_loc == nullptr && max_loc == nullptr);
-
-    _input         = input;
-    _min_max_count = min_max_count;
-
-    // Set kernel build options
-    std::set<std::string> build_opts;
-    build_opts.emplace("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
-    build_opts.emplace((min_max_count != nullptr) ? "-DCOUNT_MIN_MAX" : "");
-    build_opts.emplace((min_loc != nullptr) ? "-DLOCATE_MIN" : "");
-    build_opts.emplace((max_loc != nullptr) ? "-DLOCATE_MAX" : "");
-    if(input->info()->data_type() == DataType::F32)
-    {
-        build_opts.emplace("-DIS_DATA_TYPE_FLOAT");
-    }
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "minmaxloc", build_opts);
-
-    // Set static arguments
-    unsigned int idx = num_arguments_per_2D_tensor(); //Skip the input and output parameters
-    _kernel.setArg(idx++, *min_max);
-    _kernel.setArg(idx++, *min_max_count);
-    if(min_loc != nullptr)
-    {
-        _kernel.setArg(idx++, min_loc->cl_buffer());
-        _kernel.setArg<cl_uint>(idx++, min_loc->max_num_values());
-    }
-    if(max_loc != nullptr)
-    {
-        _kernel.setArg(idx++, max_loc->cl_buffer());
-        _kernel.setArg<cl_uint>(idx++, max_loc->max_num_values());
-    }
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 1;
-    Window                 win                               = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration));
-    update_window_and_padding(win, AccessWindowHorizontal(input->info(), 0, num_elems_processed_per_iteration));
-    ICLKernel::configure_internal(win);
-}
-
-void CLMinMaxLocationKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
-
-    static const unsigned int zero_count = 0;
-    queue.enqueueWriteBuffer(*_min_max_count, CL_FALSE, 0 * sizeof(zero_count), sizeof(zero_count), &zero_count);
-    queue.enqueueWriteBuffer(*_min_max_count, CL_FALSE, 1 * sizeof(zero_count), sizeof(zero_count), &zero_count);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLMinMaxLocationKernel.h b/src/core/CL/kernels/CLMinMaxLocationKernel.h
deleted file mode 100644
index 2196abe033..0000000000
--- a/src/core/CL/kernels/CLMinMaxLocationKernel.h
+++ /dev/null
@@ -1,124 +0,0 @@
-/*
- * Copyright (c) 2017-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLMINMAXLOCATIONKERNEL_H
-#define ARM_COMPUTE_CLMINMAXLOCATIONKERNEL_H
-
-#include "arm_compute/core/CL/ICLArray.h"
-#include "src/core/CL/ICLKernel.h"
-
-#include <array>
-
-namespace arm_compute
-{
-class ICLTensor;
-using ICLImage = ICLTensor;
-
-/** Interface for the kernel to perform min max search on an image.
- */
-class CLMinMaxKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLMinMaxKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLMinMaxKernel(const CLMinMaxKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLMinMaxKernel &operator=(const CLMinMaxKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLMinMaxKernel(CLMinMaxKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLMinMaxKernel &operator=(CLMinMaxKernel &&) = default;
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in]  input   Input Image. Data types supported: U8/S16/F32.
-     * @param[out] min_max Buffer of 2 elements to store the min value at position 0 and the max value at position 1. Data type supported: S32 if input type is U8/S16, F32 if input type is F32.
-     */
-    void configure(const ICLImage *input, cl::Buffer *min_max);
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Input Image. Data types supported: U8/S16/F32.
-     * @param[out] min_max         Buffer of 2 elements to store the min value at position 0 and the max value at position 1. Data type supported: S32 if input type is U8/S16, F32 if input type is F32.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLImage *input, cl::Buffer *min_max);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_input;               /**< Input image. */
-    cl::Buffer      *_min_max;             /**< Minimum/maximum value. */
-    std::array<int, 2> _data_type_max_min; /**< Maximum and minimum data type value respectively. */
-};
-
-/** Interface for the kernel to find min max locations of an image.
- */
-class CLMinMaxLocationKernel : public ICLKernel
-{
-public:
-    /** Constructor */
-    CLMinMaxLocationKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLMinMaxLocationKernel(const CLMinMaxLocationKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLMinMaxLocationKernel &operator=(const CLMinMaxLocationKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLMinMaxLocationKernel(CLMinMaxLocationKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLMinMaxLocationKernel &operator=(CLMinMaxLocationKernel &&) = default;
-    /** Initialise the kernel's input and outputs.
-     *
-     * @note When locations of min and max occurrences are requested, the reported number of locations is limited to the given array size.
-     *
-     * @param[in]  input         Input image. Data types supported: U8/S16/F32.
-     * @param[out] min_max       Buffer of 2 elements to store the min value at position 0 and the max value at position 1. Data type supported: S32 if input type is U8/S16, F32 if input type is F32.
-     * @param[out] min_max_count Buffer of 2 elements to store the min value occurrences at position 0 and the max value occurrences at position 1. Data type supported: S32
-     * @param[out] min_loc       (Optional) Array of Coordinates2D used to store minimum value locations.
-     * @param[out] max_loc       (Optional) Array of Coordinates2D used to store maximum value locations.
-     */
-    void configure(const ICLImage *input, cl::Buffer *min_max, cl::Buffer *min_max_count,
-                   ICLCoordinates2DArray *min_loc = nullptr, ICLCoordinates2DArray *max_loc = nullptr);
-    /** Initialise the kernel's input and outputs.
-     *
-     * @note When locations of min and max occurrences are requested, the reported number of locations is limited to the given array size.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Input image. Data types supported: U8/S16/F32.
-     * @param[out] min_max         Buffer of 2 elements to store the min value at position 0 and the max value at position 1. Data type supported: S32 if input type is U8/S16, F32 if input type is F32.
-     * @param[out] min_max_count   Buffer of 2 elements to store the min value occurrences at position 0 and the max value occurrences at position 1. Data type supported: S32
-     * @param[out] min_loc         (Optional) Array of Coordinates2D used to store minimum value locations.
-     * @param[out] max_loc         (Optional) Array of Coordinates2D used to store maximum value locations.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLImage *input, cl::Buffer *min_max, cl::Buffer *min_max_count,
-                   ICLCoordinates2DArray *min_loc = nullptr, ICLCoordinates2DArray *max_loc = nullptr);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLImage *_input;         /**< Input image. */
-    cl::Buffer     *_min_max_count; /**< Minimum/maximum value occurrences. */
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLMINMAXLOCATIONKERNEL_H */
diff --git a/src/core/CL/kernels/CLNonLinearFilterKernel.cpp b/src/core/CL/kernels/CLNonLinearFilterKernel.cpp
deleted file mode 100644
index c73acaf1d8..0000000000
--- a/src/core/CL/kernels/CLNonLinearFilterKernel.cpp
+++ /dev/null
@@ -1,104 +0,0 @@
-/*
- * Copyright (c) 2017-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLNonLinearFilterKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-
-#include <algorithm>
-#include <cmath>
-#include <cstdlib>
-#include <set>
-#include <sstream>
-#include <string>
-
-using namespace arm_compute;
-
-CLNonLinearFilterKernel::CLNonLinearFilterKernel()
-    : _border_size(0)
-{
-}
-
-BorderSize CLNonLinearFilterKernel::border_size() const
-{
-    return _border_size;
-}
-
-void CLNonLinearFilterKernel::configure(const ICLTensor *input, ICLTensor *output, NonLinearFilterFunction function,
-                                        unsigned int mask_size, MatrixPattern pattern, const uint8_t *mask,
-                                        bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, function, mask_size, pattern, mask, border_undefined);
-}
-
-void CLNonLinearFilterKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, NonLinearFilterFunction function,
-                                        unsigned int mask_size, MatrixPattern pattern, const uint8_t *mask,
-                                        bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON(mask_size != 3 && mask_size != 5);
-    ARM_COMPUTE_ERROR_ON_MSG(pattern == MatrixPattern::OTHER, "MatrixPattern::OTHER is not supported!");
-    ARM_COMPUTE_UNUSED(mask);
-
-    _input       = input;
-    _output      = output;
-    _border_size = BorderSize(mask_size / 2);
-
-    // Define build options
-    std::set<std::string> build_opts;
-    build_opts.emplace("-D" + string_from_non_linear_filter_function(function));
-
-    // Define kernel
-    std::string pattern_name = string_from_matrix_pattern(pattern);
-    std::transform(pattern_name.begin(), pattern_name.end(), pattern_name.begin(), ::tolower);
-    std::stringstream ss;
-    ss << "non_linear_filter_" << pattern_name << mask_size << "x" << mask_size;
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, ss.str(), build_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    const unsigned int     num_rows_read_per_iteration       = mask_size;
-
-    Window win = calculate_max_window(*input->info(), num_elems_processed_per_iteration, border_undefined, border_size());
-
-    AccessWindowRectangle  input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-}
diff --git a/src/core/CL/kernels/CLNonLinearFilterKernel.h b/src/core/CL/kernels/CLNonLinearFilterKernel.h
deleted file mode 100644
index ed42063d2b..0000000000
--- a/src/core/CL/kernels/CLNonLinearFilterKernel.h
+++ /dev/null
@@ -1,77 +0,0 @@
-/*
- * Copyright (c) 2017-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLNONLINEARFILTERKERNEL_H
-#define ARM_COMPUTE_CLNONLINEARFILTERKERNEL_H
-
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-#include <cstdint>
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the kernel to apply a non-linear filter */
-class CLNonLinearFilterKernel : public ICLSimple2DKernel
-{
-public:
-    /** Default constructor */
-    CLNonLinearFilterKernel();
-    /** Set the source, destination and border mode of the kernel
-     *
-     * @param[in]  input            Source tensor. Data types supported: U8
-     * @param[out] output           Destination tensor. Data types supported: U8
-     * @param[in]  function         Non linear function to perform
-     * @param[in]  mask_size        Mask size. Supported sizes: 3, 5
-     * @param[in]  pattern          Mask pattern
-     * @param[in]  mask             The given mask. Will be used only if pattern is specified to PATTERN_OTHER
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, NonLinearFilterFunction function,
-                   unsigned int mask_size, MatrixPattern pattern, const uint8_t *mask,
-                   bool border_undefined);
-    /** Set the source, destination and border mode of the kernel
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            Source tensor. Data types supported: U8
-     * @param[out] output           Destination tensor. Data types supported: U8
-     * @param[in]  function         Non linear function to perform
-     * @param[in]  mask_size        Mask size. Supported sizes: 3, 5
-     * @param[in]  pattern          Mask pattern
-     * @param[in]  mask             The given mask. Will be used only if pattern is specified to PATTERN_OTHER
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, NonLinearFilterFunction function,
-                   unsigned int mask_size, MatrixPattern pattern, const uint8_t *mask,
-                   bool border_undefined);
-
-    // Inherited methods overridden:
-    BorderSize border_size() const override;
-
-private:
-    BorderSize _border_size; /**< Border size */
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLNONLINEARFILTERKERNEL_H */
diff --git a/src/core/CL/kernels/CLNonMaximaSuppression3x3Kernel.cpp b/src/core/CL/kernels/CLNonMaximaSuppression3x3Kernel.cpp
deleted file mode 100644
index 7d5c5ba7e1..0000000000
--- a/src/core/CL/kernels/CLNonMaximaSuppression3x3Kernel.cpp
+++ /dev/null
@@ -1,78 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLNonMaximaSuppression3x3Kernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Types.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-
-#include <set>
-#include <string>
-
-using namespace arm_compute;
-
-BorderSize CLNonMaximaSuppression3x3Kernel::border_size() const
-{
-    return BorderSize(1);
-}
-
-void CLNonMaximaSuppression3x3Kernel::configure(const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, border_undefined);
-}
-
-void CLNonMaximaSuppression3x3Kernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::F32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8, DataType::F32);
-
-    _input  = input;
-    _output = output;
-
-    // Create kernel
-    std::set<std::string> build_opts = { ("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type())) };
-    _kernel                          = create_kernel(compile_context, "non_max_suppression", build_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    constexpr unsigned int num_rows_read_per_iteration       = 3;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-}
diff --git a/src/core/CL/kernels/CLNonMaximaSuppression3x3Kernel.h b/src/core/CL/kernels/CLNonMaximaSuppression3x3Kernel.h
deleted file mode 100644
index d9ed60ce6b..0000000000
--- a/src/core/CL/kernels/CLNonMaximaSuppression3x3Kernel.h
+++ /dev/null
@@ -1,60 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLNONMAXIMASUPPRESSION3x3KERNEL_H
-#define ARM_COMPUTE_CLNONMAXIMASUPPRESSION3x3KERNEL_H
-
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface to perform Non-Maxima suppression over a 3x3 window using OpenCL
- *
- * @note Used by @ref CLFastCorners and @ref CLHarrisCorners
- */
-class CLNonMaximaSuppression3x3Kernel : public ICLSimple2DKernel
-{
-public:
-    /** Initialise the kernel's sources, destinations and border mode.
-     *
-     * @param[in]  input            Source tensor. Data types supported: U8, F32. (Must be the same as the output tensor)
-     * @param[out] output           Destination tensor. Data types supported: U8, F32. (Must be the same as the input tensor)
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, bool border_undefined);
-    /** Initialise the kernel's sources, destinations and border mode.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            Source tensor. Data types supported: U8, F32. (Must be the same as the output tensor)
-     * @param[out] output           Destination tensor. Data types supported: U8, F32. (Must be the same as the input tensor)
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, bool border_undefined);
-
-    // Inherited methods overridden:
-    BorderSize border_size() const override;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLNONMAXIMASUPPRESSION3x3KERNEL_H */
diff --git a/src/core/CL/kernels/CLScharr3x3Kernel.cpp b/src/core/CL/kernels/CLScharr3x3Kernel.cpp
deleted file mode 100644
index 7ceddc9626..0000000000
--- a/src/core/CL/kernels/CLScharr3x3Kernel.cpp
+++ /dev/null
@@ -1,127 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLScharr3x3Kernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-
-#include <set>
-#include <string>
-
-using namespace arm_compute;
-
-CLScharr3x3Kernel::CLScharr3x3Kernel()
-    : _run_scharr_x(false), _run_scharr_y(false), _input(nullptr), _output_x(nullptr), _output_y(nullptr)
-{
-}
-
-BorderSize CLScharr3x3Kernel::border_size() const
-{
-    return BorderSize(1);
-}
-
-void CLScharr3x3Kernel::configure(const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output_x, output_y, border_undefined);
-}
-
-void CLScharr3x3Kernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON((output_x == nullptr) && (output_y == nullptr));
-
-    _run_scharr_x = output_x != nullptr;
-    _run_scharr_y = output_y != nullptr;
-
-    if(_run_scharr_x)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_x, 1, DataType::S16);
-    }
-
-    if(_run_scharr_y)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_y, 1, DataType::S16);
-    }
-
-    _input    = input;
-    _output_x = output_x;
-    _output_y = output_y;
-
-    // Set build options
-    std::set<std::string> build_opts;
-
-    if(_run_scharr_x)
-    {
-        build_opts.insert("-DGRAD_X");
-    }
-
-    if(_run_scharr_y)
-    {
-        build_opts.insert("-DGRAD_Y");
-    }
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "scharr3x3", build_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-    constexpr unsigned int num_rows_read_per_iteration       = 3;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_x_access(output_x == nullptr ? nullptr : output_x->info(), 0, num_elems_written_per_iteration);
-    AccessWindowHorizontal output_y_access(output_y == nullptr ? nullptr : output_y->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_access, output_x_access, output_y_access);
-
-    output_x_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-    output_y_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-}
-
-void CLScharr3x3Kernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-        add_2D_tensor_argument_if((_run_scharr_x), idx, _output_x, slice);
-        add_2D_tensor_argument_if((_run_scharr_y), idx, _output_y, slice);
-
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLScharr3x3Kernel.h b/src/core/CL/kernels/CLScharr3x3Kernel.h
deleted file mode 100644
index a670da5b6f..0000000000
--- a/src/core/CL/kernels/CLScharr3x3Kernel.h
+++ /dev/null
@@ -1,97 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLSCHARR3X3KERNEL_H
-#define ARM_COMPUTE_CLSCHARR3X3KERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the kernel to run a 3x3 Scharr filter on a tensor.
- *
- * @f[
- *      \mathbf{G}_x=\begin{vmatrix}
- *      -3 & 0 & +3\\
- *      -10& 0 & +10\\
- *      -3 & 0 & +3
- *      \end{vmatrix}
- * @f]
- * @f[
- *      \mathbf{G}_y=\begin{vmatrix}
- *      -3 & -10 & -3\\
- *       0 & 0 & 0\\
- *      +3 & +10 & +3
- *      \end{vmatrix}
- * @f]
- */
-class CLScharr3x3Kernel : public ICLKernel
-{
-public:
-    /** Default constructor: initialize all the pointers to nullptr and parameters to zero. */
-    CLScharr3x3Kernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLScharr3x3Kernel(const CLScharr3x3Kernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLScharr3x3Kernel &operator=(const CLScharr3x3Kernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLScharr3x3Kernel(CLScharr3x3Kernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLScharr3x3Kernel &operator=(CLScharr3x3Kernel &&) = default;
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @note At least one of output_x or output_y must be set.
-     *
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S16.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S16.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @note At least one of output_x or output_y must be set.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S16.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S16.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-    BorderSize border_size() const override;
-
-private:
-    bool             _run_scharr_x; /**< Do we need to run Scharr X ? */
-    bool             _run_scharr_y; /**< Do we need to run Scharr Y ? */
-    const ICLTensor *_input;        /**< Input image */
-    ICLTensor       *_output_x;     /**< Output image for scharr X */
-    ICLTensor       *_output_y;     /**< Output image for scharr Y */
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLSCHARR3X3KERNEL_H */
diff --git a/src/core/CL/kernels/CLSobel3x3Kernel.cpp b/src/core/CL/kernels/CLSobel3x3Kernel.cpp
deleted file mode 100644
index a87677a38f..0000000000
--- a/src/core/CL/kernels/CLSobel3x3Kernel.cpp
+++ /dev/null
@@ -1,141 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLSobel3x3Kernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <set>
-#include <string>
-
-using namespace arm_compute;
-
-CLSobel3x3Kernel::CLSobel3x3Kernel()
-    : _input(nullptr), _output_x(nullptr), _output_y(nullptr), _run_sobel_x(false), _run_sobel_y(false)
-{
-}
-
-BorderSize CLSobel3x3Kernel::border_size() const
-{
-    return BorderSize(1);
-}
-
-void CLSobel3x3Kernel::configure(const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output_x, output_y, border_undefined);
-}
-
-void CLSobel3x3Kernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON((output_x == nullptr) && (output_y == nullptr));
-
-    _run_sobel_x = output_x != nullptr;
-    _run_sobel_y = output_y != nullptr;
-
-    if(_run_sobel_x)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_x, 1, DataType::S16);
-    }
-
-    if(_run_sobel_y)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_y, 1, DataType::S16);
-    }
-
-    _input    = input;
-    _output_x = output_x;
-    _output_y = output_y;
-
-    // Set build options
-    std::set<std::string> build_opts;
-
-    if(_run_sobel_x)
-    {
-        build_opts.insert("-DGRAD_X");
-    }
-
-    if(_run_sobel_y)
-    {
-        build_opts.insert("-DGRAD_Y");
-    }
-
-    // Create kernel
-    const std::string kernel_name = std::string("sobel3x3");
-    _kernel                       = create_kernel(compile_context, kernel_name, build_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-    constexpr unsigned int num_rows_read_per_iteration       = 3;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_x_access(output_x == nullptr ? nullptr : output_x->info(), 0, num_elems_written_per_iteration);
-    AccessWindowHorizontal output_y_access(output_y == nullptr ? nullptr : output_y->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_access, output_x_access, output_y_access);
-
-    output_x_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-    output_y_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(border_undefined);
-}
-
-void CLSobel3x3Kernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-        add_2D_tensor_argument_if((_run_sobel_x), idx, _output_x, slice);
-        add_2D_tensor_argument_if((_run_sobel_y), idx, _output_y, slice);
-
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLSobel3x3Kernel.h b/src/core/CL/kernels/CLSobel3x3Kernel.h
deleted file mode 100644
index fed8068762..0000000000
--- a/src/core/CL/kernels/CLSobel3x3Kernel.h
+++ /dev/null
@@ -1,83 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLSOBEL3X3KERNEL_H
-#define ARM_COMPUTE_CLSOBEL3X3KERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the kernel to run a 3x3 Sobel filter on a tensor. */
-class CLSobel3x3Kernel : public ICLKernel
-{
-public:
-    /** Default constructor: initialize all the pointers to nullptr and parameters to zero. */
-    CLSobel3x3Kernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLSobel3x3Kernel(const CLSobel3x3Kernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLSobel3x3Kernel &operator=(const CLSobel3x3Kernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLSobel3x3Kernel(CLSobel3x3Kernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLSobel3x3Kernel &operator=(CLSobel3x3Kernel &&) = default;
-    /** Default destructor */
-    ~CLSobel3x3Kernel() = default;
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @note At least one of output_x or output_y must be set.
-     *
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S16.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S16.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @note At least one of output_x or output_y must be set.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S16.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S16.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-    BorderSize border_size() const override;
-
-private:
-    const ICLTensor *_input;       /**< Input tensor */
-    ICLTensor       *_output_x;    /**< Output tensor for Sobel X */
-    ICLTensor       *_output_y;    /**< Output tensor for Sobel Y */
-    bool             _run_sobel_x; /**< Do we need to run Sobel X ? */
-    bool             _run_sobel_y; /**< Do we need to run Sobel Y ? */
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLSOBEL3X3KERNEL_H */
diff --git a/src/core/CL/kernels/CLSobel5x5Kernel.cpp b/src/core/CL/kernels/CLSobel5x5Kernel.cpp
deleted file mode 100644
index c450becd1d..0000000000
--- a/src/core/CL/kernels/CLSobel5x5Kernel.cpp
+++ /dev/null
@@ -1,251 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLSobel5x5Kernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <set>
-#include <string>
-
-using namespace arm_compute;
-
-CLSobel5x5HorKernel::CLSobel5x5HorKernel()
-    : _input(nullptr), _output_x(nullptr), _output_y(nullptr), _run_sobel_x(false), _run_sobel_y(false), _border_size(0)
-{
-}
-
-BorderSize CLSobel5x5HorKernel::border_size() const
-{
-    return _border_size;
-}
-
-void CLSobel5x5HorKernel::configure(const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output_x, output_y, border_undefined);
-}
-
-void CLSobel5x5HorKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON((output_x == nullptr) && (output_y == nullptr));
-
-    _run_sobel_x = output_x != nullptr;
-    _run_sobel_y = output_y != nullptr;
-
-    if(_run_sobel_x)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_x, 1, DataType::S16);
-    }
-
-    if(_run_sobel_y)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_y, 1, DataType::S16);
-    }
-
-    _input       = input;
-    _output_x    = output_x;
-    _output_y    = output_y;
-    _border_size = BorderSize(border_undefined ? 0 : 2, 2);
-
-    // Set build options
-    std::set<std::string> build_opts;
-
-    if(_run_sobel_x)
-    {
-        build_opts.insert("-DGRAD_X");
-    }
-
-    if(_run_sobel_y)
-    {
-        build_opts.insert("-DGRAD_Y");
-    }
-
-    // Create kernel
-    const std::string kernel_name = std::string("sobel_separable1x5");
-    _kernel                       = create_kernel(compile_context, kernel_name, build_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-
-    Window win = calculate_max_window_horizontal(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowHorizontal input_access(input->info(), -border_size().left, num_elems_read_per_iteration);
-    AccessWindowHorizontal output_x_access(output_x == nullptr ? nullptr : output_x->info(), 0, num_elems_written_per_iteration);
-    AccessWindowHorizontal output_y_access(output_y == nullptr ? nullptr : output_y->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_access, output_x_access, output_y_access);
-
-    output_x_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-    output_y_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(border_undefined);
-}
-
-void CLSobel5x5HorKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-        add_2D_tensor_argument_if((_run_sobel_x), idx, _output_x, slice);
-        add_2D_tensor_argument_if((_run_sobel_y), idx, _output_y, slice);
-
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
-
-CLSobel5x5VertKernel::CLSobel5x5VertKernel()
-    : _input_x(nullptr), _input_y(nullptr), _output_x(nullptr), _output_y(nullptr), _run_sobel_x(false), _run_sobel_y(false)
-{
-}
-
-BorderSize CLSobel5x5VertKernel::border_size() const
-{
-    return BorderSize{ 2, 0 };
-}
-
-void CLSobel5x5VertKernel::configure(const ICLTensor *input_x, const ICLTensor *input_y, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input_x, input_y, output_x, output_y, border_undefined);
-}
-
-void CLSobel5x5VertKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input_x, const ICLTensor *input_y, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON((output_x == nullptr) && (output_y == nullptr));
-
-    _run_sobel_x = output_x != nullptr;
-    _run_sobel_y = output_y != nullptr;
-
-    if(_run_sobel_x)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input_x, 1, DataType::S16);
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_x, 1, DataType::S16);
-    }
-
-    if(_run_sobel_y)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input_y, 1, DataType::S16);
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_y, 1, DataType::S16);
-    }
-
-    _input_x  = input_x;
-    _input_y  = input_y;
-    _output_x = output_x;
-    _output_y = output_y;
-
-    // Set build options
-    std::set<std::string> build_opts;
-
-    if(_run_sobel_x)
-    {
-        build_opts.insert("-DGRAD_X");
-    }
-
-    if(_run_sobel_y)
-    {
-        build_opts.insert("-DGRAD_Y");
-    }
-
-    // Create kernel
-    const std::string kernel_name = std::string("sobel_separable5x1");
-    _kernel                       = create_kernel(compile_context, kernel_name, build_opts);
-
-    const ICLTensor *input = _run_sobel_x ? _input_x : _input_y;
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 8;
-    constexpr unsigned int num_rows_read_per_iteration       = 5;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_x_access(input_x == nullptr ? nullptr : input_x->info(), 0, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowRectangle  input_y_access(input_y == nullptr ? nullptr : input_y->info(), 0, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_x_access(output_x == nullptr ? nullptr : output_x->info(), 0, num_elems_written_per_iteration);
-    AccessWindowHorizontal output_y_access(output_y == nullptr ? nullptr : output_y->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_x_access, input_y_access, output_x_access, output_y_access);
-
-    output_x_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-    output_y_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(border_undefined);
-}
-
-void CLSobel5x5VertKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument_if((_run_sobel_x), idx, _input_x, slice);
-        add_2D_tensor_argument_if((_run_sobel_x), idx, _output_x, slice);
-        add_2D_tensor_argument_if((_run_sobel_y), idx, _input_y, slice);
-        add_2D_tensor_argument_if((_run_sobel_y), idx, _output_y, slice);
-
-        _kernel.setArg(idx++, 0 /*dummy*/);
-
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLSobel5x5Kernel.h b/src/core/CL/kernels/CLSobel5x5Kernel.h
deleted file mode 100644
index a163ac932a..0000000000
--- a/src/core/CL/kernels/CLSobel5x5Kernel.h
+++ /dev/null
@@ -1,139 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLSOBEL5X5KERNEL_H
-#define ARM_COMPUTE_CLSOBEL5X5KERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the kernel to run the horizontal pass of 5x5 Sobel filter on a tensor. */
-class CLSobel5x5HorKernel : public ICLKernel
-{
-public:
-    /** Default constructor: initialize all the pointers to nullptr and parameters to zero. */
-    CLSobel5x5HorKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLSobel5x5HorKernel(const CLSobel5x5HorKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLSobel5x5HorKernel &operator=(const CLSobel5x5HorKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLSobel5x5HorKernel(CLSobel5x5HorKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLSobel5x5HorKernel &operator=(CLSobel5x5HorKernel &&) = default;
-    /** Default destructor */
-    ~CLSobel5x5HorKernel() = default;
-
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @note At least one of output_x or output_y must be set.
-     *
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S16.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S16.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @note At least one of output_x or output_y must be set.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S16.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S16.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-    BorderSize border_size() const override;
-
-private:
-    const ICLTensor *_input;       /**< Input tensor */
-    ICLTensor       *_output_x;    /**< X output of horizontal pass */
-    ICLTensor       *_output_y;    /**< Y output of horizontal pass */
-    bool             _run_sobel_x; /**< Do we need to run Sobel X ? */
-    bool             _run_sobel_y; /**< Do we need to run Sobel Y ? */
-    BorderSize       _border_size; /**< Border size */
-};
-
-/** Interface for the kernel to run the vertical pass of 5x5 Sobel filter on a tensor. */
-class CLSobel5x5VertKernel : public ICLKernel
-{
-public:
-    /** Default constructor: initialize all the pointers to nullptr and parameters to zero. */
-    CLSobel5x5VertKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLSobel5x5VertKernel(const CLSobel5x5VertKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLSobel5x5VertKernel &operator=(const CLSobel5x5VertKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLSobel5x5VertKernel(CLSobel5x5VertKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLSobel5x5VertKernel &operator=(CLSobel5x5VertKernel &&) = default;
-    /** Default destructor */
-    ~CLSobel5x5VertKernel() = default;
-
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @note At least one of output_x or output_y must be set and the corresponding input.
-     *
-     * @param[in]  input_x          (Optional) Input for X (X output of horizontal pass). Data types supported: S16.
-     * @param[in]  input_y          (Optional) Input for Y (Y output of horizontal pass). Data types supported: S16.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S16.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S16.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input_x, const ICLTensor *input_y, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @note At least one of output_x or output_y must be set and the corresponding input.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input_x          (Optional) Input for X (X output of horizontal pass). Data types supported: S16.
-     * @param[in]  input_y          (Optional) Input for Y (Y output of horizontal pass). Data types supported: S16.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S16.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S16.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input_x, const ICLTensor *input_y, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-    BorderSize border_size() const override;
-
-private:
-    const ICLTensor *_input_x;     /**< X input (X output of the horizontal pass) */
-    const ICLTensor *_input_y;     /**< Y input (Y output of the horizontal pass) */
-    ICLTensor       *_output_x;    /**< X output of sobel */
-    ICLTensor       *_output_y;    /**< Y output of sobel */
-    bool             _run_sobel_x; /**< Do we need to run sobel X? */
-    bool             _run_sobel_y; /**< Do we need to run sobel Y? */
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLSOBEL5X5KERNEL_H */
diff --git a/src/core/CL/kernels/CLSobel7x7Kernel.cpp b/src/core/CL/kernels/CLSobel7x7Kernel.cpp
deleted file mode 100644
index 1cfa74f7b3..0000000000
--- a/src/core/CL/kernels/CLSobel7x7Kernel.cpp
+++ /dev/null
@@ -1,255 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLSobel7x7Kernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <set>
-#include <string>
-
-using namespace arm_compute;
-
-CLSobel7x7HorKernel::CLSobel7x7HorKernel()
-    : _input(nullptr), _output_x(nullptr), _output_y(nullptr), _run_sobel_x(false), _run_sobel_y(false), _border_size(0)
-{
-}
-
-BorderSize CLSobel7x7HorKernel::border_size() const
-{
-    return _border_size;
-}
-
-void CLSobel7x7HorKernel::configure(const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output_x, output_y, border_undefined);
-}
-
-void CLSobel7x7HorKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON((output_x == nullptr) && (output_y == nullptr));
-
-    _run_sobel_x = output_x != nullptr;
-    _run_sobel_y = output_y != nullptr;
-
-    if(_run_sobel_x)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_x, 1, DataType::S32);
-    }
-
-    if(_run_sobel_y)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_y, 1, DataType::S32);
-    }
-
-    _input       = input;
-    _output_x    = output_x;
-    _output_y    = output_y;
-    _border_size = BorderSize(border_undefined ? 0 : 3, 3);
-
-    // Construct kernel name
-    const std::string kernel_name = "sobel_separable1x7";
-
-    // Set build options
-    std::set<std::string> build_opts;
-
-    if(_run_sobel_x)
-    {
-        build_opts.insert("-DGRAD_X");
-    }
-
-    if(_run_sobel_y)
-    {
-        build_opts.insert("-DGRAD_Y");
-    }
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, kernel_name, build_opts);
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 16;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-
-    Window win = calculate_max_window_horizontal(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowHorizontal input_access(input->info(), -border_size().left, num_elems_read_per_iteration);
-    AccessWindowHorizontal output_x_access(output_x == nullptr ? nullptr : output_x->info(), 0, num_elems_written_per_iteration);
-    AccessWindowHorizontal output_y_access(output_y == nullptr ? nullptr : output_y->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_access, output_x_access, output_y_access);
-
-    output_x_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-    output_y_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(border_undefined);
-}
-
-void CLSobel7x7HorKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-    do
-    {
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input, slice);
-        add_2D_tensor_argument_if((_run_sobel_x), idx, _output_x, slice);
-        add_2D_tensor_argument_if((_run_sobel_y), idx, _output_y, slice);
-
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
-
-CLSobel7x7VertKernel::CLSobel7x7VertKernel()
-    : _input_x(nullptr), _input_y(nullptr), _output_x(nullptr), _output_y(nullptr), _run_sobel_x(false), _run_sobel_y(false)
-{
-}
-
-BorderSize CLSobel7x7VertKernel::border_size() const
-{
-    return BorderSize{ 3, 0 };
-}
-
-void CLSobel7x7VertKernel::configure(const ICLTensor *input_x, const ICLTensor *input_y, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input_x, input_y, output_x, output_y, border_undefined);
-}
-
-void CLSobel7x7VertKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input_x, const ICLTensor *input_y, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
-{
-    ARM_COMPUTE_ERROR_ON((output_x == nullptr) && (output_y == nullptr));
-
-    _run_sobel_x = output_x != nullptr;
-    _run_sobel_y = output_y != nullptr;
-
-    if(_run_sobel_x)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input_x, 1, DataType::S32);
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_x, 1, DataType::S32);
-    }
-
-    if(_run_sobel_y)
-    {
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input_y, 1, DataType::S32);
-        ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_y, 1, DataType::S32);
-    }
-
-    _input_x  = input_x;
-    _input_y  = input_y;
-    _output_x = output_x;
-    _output_y = output_y;
-
-    // Set build options
-    std::set<std::string> build_opts;
-
-    if(_run_sobel_x)
-    {
-        build_opts.insert("-DGRAD_X");
-    }
-
-    if(_run_sobel_y)
-    {
-        build_opts.insert("-DGRAD_Y");
-    }
-
-    // Create kernel
-    const std::string kernel_name = std::string("sobel_separable7x1");
-    _kernel                       = create_kernel(compile_context, kernel_name, build_opts);
-
-    const ICLTensor *input = _run_sobel_x ? _input_x : _input_y;
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    constexpr unsigned int num_elems_written_per_iteration   = 8;
-    constexpr unsigned int num_elems_read_per_iteration      = 8;
-    constexpr unsigned int num_rows_read_per_iteration       = 7;
-
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
-
-    AccessWindowRectangle  input_x_access(input_x == nullptr ? nullptr : input_x->info(), 0, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowRectangle  input_y_access(input_y == nullptr ? nullptr : input_y->info(), 0, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
-    AccessWindowHorizontal output_x_access(output_x == nullptr ? nullptr : output_x->info(), 0, num_elems_written_per_iteration);
-    AccessWindowHorizontal output_y_access(output_y == nullptr ? nullptr : output_y->info(), 0, num_elems_written_per_iteration);
-
-    update_window_and_padding(win, input_x_access, input_y_access, output_x_access, output_y_access);
-
-    output_x_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-    output_y_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(border_undefined);
-}
-
-void CLSobel7x7VertKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_2D();
-
-    do
-    {
-        unsigned int idx = 0;
-
-        add_2D_tensor_argument_if((_run_sobel_x), idx, _input_x, slice);
-        add_2D_tensor_argument_if((_run_sobel_x), idx, _output_x, slice);
-        add_2D_tensor_argument_if((_run_sobel_y), idx, _input_y, slice);
-        add_2D_tensor_argument_if((_run_sobel_y), idx, _output_y, slice);
-
-        _kernel.setArg(idx++, 0 /*dummy*/);
-
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_2D(slice));
-}
diff --git a/src/core/CL/kernels/CLSobel7x7Kernel.h b/src/core/CL/kernels/CLSobel7x7Kernel.h
deleted file mode 100644
index c85f0aedf9..0000000000
--- a/src/core/CL/kernels/CLSobel7x7Kernel.h
+++ /dev/null
@@ -1,139 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLSOBEL7X7KERNEL_H
-#define ARM_COMPUTE_CLSOBEL7X7KERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the kernel to run the horizontal pass of 7x7 Sobel filter on a tensor. */
-class CLSobel7x7HorKernel : public ICLKernel
-{
-public:
-    /** Default constructor: initialize all the pointers to nullptr and parameters to zero. */
-    CLSobel7x7HorKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLSobel7x7HorKernel(const CLSobel7x7HorKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLSobel7x7HorKernel &operator=(const CLSobel7x7HorKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLSobel7x7HorKernel(CLSobel7x7HorKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLSobel7x7HorKernel &operator=(CLSobel7x7HorKernel &&) = default;
-    /** Default destructor */
-    ~CLSobel7x7HorKernel() = default;
-
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @note At least one of output_x or output_y must be set.
-     *
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S32.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S32.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @note At least one of output_x or output_y must be set.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input            Source tensor. Data types supported: U8.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S32.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S32.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-    BorderSize border_size() const override;
-
-private:
-    const ICLTensor *_input;       /**< Input tensor */
-    ICLTensor       *_output_x;    /**< X output of horizontal pass */
-    ICLTensor       *_output_y;    /**< Y output of horizontal pass */
-    bool             _run_sobel_x; /**< Do we need to run Sobel X ? */
-    bool             _run_sobel_y; /**< Do we need to run Sobel Y ? */
-    BorderSize       _border_size; /**< Border size */
-};
-
-/** Interface for the kernel to run the vertical pass of 7x7 Sobel filter on a tensor. */
-class CLSobel7x7VertKernel : public ICLKernel
-{
-public:
-    /** Default constructor: initialize all the pointers to nullptr and parameters to zero. */
-    CLSobel7x7VertKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLSobel7x7VertKernel(const CLSobel7x7VertKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLSobel7x7VertKernel &operator=(const CLSobel7x7VertKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLSobel7x7VertKernel(CLSobel7x7VertKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLSobel7x7VertKernel &operator=(CLSobel7x7VertKernel &&) = default;
-    /** Default destructor */
-    ~CLSobel7x7VertKernel() = default;
-
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @note At least one of output_x or output_y must be set and the corresponding input.
-     *
-     * @param[in]  input_x          (Optional) Input for X (X output of horizontal pass). Data types supported: S32.
-     * @param[in]  input_y          (Optional) Input for Y (Y output of horizontal pass). Data types supported: S32.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S32.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S32.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const ICLTensor *input_x, const ICLTensor *input_y, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-    /** Initialise the kernel's source, destination and border.
-     *
-     * @note At least one of output_x or output_y must be set and the corresponding input.
-     *
-     * @param[in]  compile_context  The compile context to be used.
-     * @param[in]  input_x          (Optional) Input for X (X output of horizontal pass). Data types supported: S32.
-     * @param[in]  input_y          (Optional) Input for Y (Y output of horizontal pass). Data types supported: S32.
-     * @param[out] output_x         (Optional) Destination tensor for the X gradient, Data types supported: S32.
-     * @param[out] output_y         (Optional) Destination tensor for the Y gradient, Data types supported: S32.
-     * @param[in]  border_undefined True if the border mode is undefined. False if it's replicate or constant.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input_x, const ICLTensor *input_y, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-    BorderSize border_size() const override;
-
-private:
-    const ICLTensor *_input_x;     /**< X input (X output of the horizontal pass) */
-    const ICLTensor *_input_y;     /**< Y input (Y output of the horizontal pass) */
-    ICLTensor       *_output_x;    /**< X output of sobel */
-    ICLTensor       *_output_y;    /**< Y output of sobel */
-    bool             _run_sobel_x; /**< Do we need to run sobel X? */
-    bool             _run_sobel_y; /**< Do we need to run sobel Y? */
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLSOBEL7X7KERNEL_H */
diff --git a/src/core/CL/kernels/CLTableLookupKernel.cpp b/src/core/CL/kernels/CLTableLookupKernel.cpp
deleted file mode 100644
index b82f4c9889..0000000000
--- a/src/core/CL/kernels/CLTableLookupKernel.cpp
+++ /dev/null
@@ -1,68 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLTableLookupKernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLLut.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Error.h"
-#include "arm_compute/core/Types.h"
-#include "arm_compute/core/Validate.h"
-
-#include <cstdint>
-#include <string>
-
-using namespace arm_compute;
-
-void CLTableLookupKernel::configure(const ICLTensor *input, const ICLLut *lut, ICLTensor *output)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, lut, output);
-}
-
-void CLTableLookupKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, const ICLLut *lut, ICLTensor *output)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S16);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8, DataType::S16);
-    ARM_COMPUTE_ERROR_ON(lut == nullptr);
-    ARM_COMPUTE_ERROR_ON(DataType::U8 != lut->type() && DataType::S16 != lut->type());
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
-
-    // Create kernel
-    std::string kernel_name = (DataType::S16 == lut->type()) ? "tablelookup_S16" : "tablelookup_U8";
-    _kernel                 = create_kernel(compile_context, kernel_name);
-
-    // Set lut argument
-    unsigned int idx = 2 * num_arguments_per_2D_tensor(); //Skip the input and output parameters
-    _kernel.setArg(idx++, lut->cl_buffer());
-    if(DataType::S16 == lut->type())
-    {
-        _kernel.setArg(idx++, lut->index_offset());
-        _kernel.setArg(idx++, static_cast<uint32_t>(lut->num_elements()));
-    }
-
-    // Configure kernel
-    constexpr unsigned int num_elems_processed_per_iteration = 8;
-    ICLSimple2DKernel::configure(input, output, num_elems_processed_per_iteration);
-}
diff --git a/src/core/CL/kernels/CLTableLookupKernel.h b/src/core/CL/kernels/CLTableLookupKernel.h
deleted file mode 100644
index c8d15cbee2..0000000000
--- a/src/core/CL/kernels/CLTableLookupKernel.h
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLTABLELOOKUPKERNEL_H
-#define ARM_COMPUTE_CLTABLELOOKUPKERNEL_H
-
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-class ICLLut;
-
-/** Interface for the kernel to perform table lookup calculations. */
-class CLTableLookupKernel : public ICLSimple2DKernel
-{
-public:
-    /** Initialise the kernel's input, lut and output.
-     *
-     * @param[in]  input  An input tensor. Data types supported: U8, S16.
-     * @param[in]  lut    The input LUT. Data types supported: U8, S16.
-     * @param[out] output The output tensor. Data types supported: U8, S16.
-     */
-    void configure(const ICLTensor *input, const ICLLut *lut, ICLTensor *output);
-    /** Initialise the kernel's input, lut and output.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           An input tensor. Data types supported: U8, S16.
-     * @param[in]  lut             The input LUT. Data types supported: U8, S16.
-     * @param[out] output          The output tensor. Data types supported: U8, S16.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, const ICLLut *lut, ICLTensor *output);
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLTABLELOOKUPKERNEL_H */
diff --git a/src/core/CL/kernels/CLThresholdKernel.cpp b/src/core/CL/kernels/CLThresholdKernel.cpp
deleted file mode 100644
index 72c22f043c..0000000000
--- a/src/core/CL/kernels/CLThresholdKernel.cpp
+++ /dev/null
@@ -1,81 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLThresholdKernel.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Validate.h"
-#include "arm_compute/core/Window.h"
-
-#include <string>
-
-namespace arm_compute
-{
-void CLThresholdKernel::configure(const ICLTensor *input, ICLTensor *output, const ThresholdKernelInfo &info)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, info);
-}
-
-void CLThresholdKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const ThresholdKernelInfo &info)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8);
-
-    // Construct kernel name
-    std::string kernel_name = "threshold";
-
-    switch(info.type)
-    {
-        case ThresholdType::BINARY:
-            kernel_name += "_binary";
-            break;
-        case ThresholdType::RANGE:
-            kernel_name += "_range";
-            break;
-        default:
-            ARM_COMPUTE_ERROR("Thresholding type not recognized");
-            break;
-    }
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, kernel_name);
-
-    // Set arguments
-    unsigned int idx = 2 * num_arguments_per_2D_tensor(); //Skip the input and output parameters
-    _kernel.setArg(idx++, info.false_value);
-    _kernel.setArg(idx++, info.true_value);
-    _kernel.setArg(idx++, info.threshold);
-
-    if(ThresholdType::RANGE == info.type)
-    {
-        _kernel.setArg(idx++, info.upper);
-    }
-
-    // Make sure _kernel is initialized before calling the parent's configure
-    constexpr unsigned int num_elems_processed_per_iteration = 16;
-    ICLSimple2DKernel::configure(input, output, num_elems_processed_per_iteration);
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLThresholdKernel.h b/src/core/CL/kernels/CLThresholdKernel.h
deleted file mode 100644
index 511eaed1bf..0000000000
--- a/src/core/CL/kernels/CLThresholdKernel.h
+++ /dev/null
@@ -1,57 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLTHRESHOLDKERNEL_H
-#define ARM_COMPUTE_CLTHRESHOLDKERNEL_H
-
-#include "arm_compute/core/KernelDescriptors.h"
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-namespace arm_compute
-{
-// Forward declarations
-class ICLTensor;
-
-/** Interface for the thresholding kernel. */
-class CLThresholdKernel : public ICLSimple2DKernel
-{
-public:
-    /**Initialise the kernel's input, output and threshold parameters.
-     *
-     * @param[in]  input  An input tensor. Data types supported: U8
-     * @param[out] output The output tensor. Data types supported: U8.
-     * @param[in]  info   Threshold descriptor
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, const ThresholdKernelInfo &info);
-    /**Initialise the kernel's input, output and threshold parameters.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           An input tensor. Data types supported: U8
-     * @param[out] output          The output tensor. Data types supported: U8.
-     * @param[in]  info            Threshold descriptor
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const ThresholdKernelInfo &info);
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_NETHRESHOLDKERNEL_H */
diff --git a/src/core/CL/kernels/CLWarpAffineKernel.cpp b/src/core/CL/kernels/CLWarpAffineKernel.cpp
deleted file mode 100644
index 600c67a528..0000000000
--- a/src/core/CL/kernels/CLWarpAffineKernel.cpp
+++ /dev/null
@@ -1,133 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLWarpAffineKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Error.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/AccessWindowStatic.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-#include <cstddef>
-#include <set>
-#include <sstream>
-#include <string>
-
-namespace arm_compute
-{
-namespace
-{
-void options_add_matrix(std::set<std::string> &options, const std::array<float, 9> &matrix)
-{
-    for(size_t i = 0; i < 6; ++i)
-    {
-        std::stringstream mat_str;
-        mat_str << "-DMAT" << i << "=" << matrix[i] << " ";
-        options.insert(mat_str.str());
-    }
-}
-} // namespace
-
-BorderSize CLWarpAffineKernel::border_size() const
-{
-    return BorderSize(1);
-}
-
-void CLWarpAffineKernel::configure(const ICLTensor *input, ICLTensor *output, const std::array<float, 9> &matrix, InterpolationPolicy policy)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, matrix, policy);
-}
-
-void CLWarpAffineKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const std::array<float, 9> &matrix, InterpolationPolicy policy)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON(InterpolationPolicy::AREA == policy);
-
-    _input  = input;
-    _output = output;
-
-    // Create build options
-    std::set<std::string> options;
-    options_add_matrix(options, matrix);
-    options.emplace(("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type())));
-
-    // Create kernel
-    std::string interpolation_name = string_from_interpolation_policy(policy);
-    std::transform(interpolation_name.begin(), interpolation_name.end(), interpolation_name.begin(), ::tolower);
-    const std::string kernel_name = "warp_affine_" + interpolation_name;
-    _kernel                       = create_kernel(compile_context, kernel_name, options);
-
-    // Set static kernel arguments
-    unsigned int idx = 2 * num_arguments_per_2D_tensor(); //Skip the input and output parameters
-    _kernel.setArg<cl_int>(idx++, input->info()->dimension(0));
-    _kernel.setArg<cl_int>(idx++, input->info()->dimension(1));
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 4;
-
-    Window win = calculate_max_window(*output->info(), Steps(num_elems_processed_per_iteration));
-
-    int       total_right  = ceil_to_multiple(input->info()->dimension(0), num_elems_processed_per_iteration);
-    const int access_right = total_right + (((total_right - input->info()->dimension(0)) == 0) ? border_size().right : 0);
-
-    AccessWindowStatic     input_access(input->info(), -border_size().left, -border_size().top, access_right, input->info()->dimension(1) + border_size().bottom);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, ValidRegion(Coordinates(), output->info()->tensor_shape()));
-
-    ICLKernel::configure_internal(win);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(2));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->info()->dimension(3));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(2));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(3));
-    _config_id += "_";
-    _config_id += lower_string(string_from_interpolation_policy(policy));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLWarpAffineKernel.h b/src/core/CL/kernels/CLWarpAffineKernel.h
deleted file mode 100644
index c600ee780d..0000000000
--- a/src/core/CL/kernels/CLWarpAffineKernel.h
+++ /dev/null
@@ -1,62 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLWARPAFFINEKERNEL_H
-#define ARM_COMPUTE_CLWARPAFFINEKERNEL_H
-
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the warp affine kernel.*/
-class CLWarpAffineKernel : public ICLSimple2DKernel
-{
-public:
-    /** Initialize the function's source, destination, interpolation policy and border_mode.
-     *
-     * @param[in]  input  Source tensor. Data types supported: U8.
-     * @param[out] output Destination tensor, Data types supported: U8.
-     * @param[in]  matrix The perspective matrix. Must be 2x3 of type float
-     *                    The matrix argument requires 9 values, the last 3 values are ignored.
-     * @param[in]  policy The interpolation type.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, const std::array<float, 9> &matrix, InterpolationPolicy policy);
-    /** Initialize the function's source, destination, interpolation policy and border_mode.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Source tensor. Data types supported: U8.
-     * @param[out] output          Destination tensor, Data types supported: U8.
-     * @param[in]  matrix          The perspective matrix. Must be 2x3 of type float
-     *                             The matrix argument requires 9 values, the last 3 values are ignored.
-     * @param[in]  policy          The interpolation type.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const std::array<float, 9> &matrix, InterpolationPolicy policy);
-
-    // Inherited methods overridden:
-    BorderSize border_size() const override;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLWARPAFFINEKERNEL_H */
diff --git a/src/core/CL/kernels/CLWarpPerspectiveKernel.cpp b/src/core/CL/kernels/CLWarpPerspectiveKernel.cpp
deleted file mode 100644
index 5f20a0bdd3..0000000000
--- a/src/core/CL/kernels/CLWarpPerspectiveKernel.cpp
+++ /dev/null
@@ -1,105 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLWarpPerspectiveKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Error.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Validate.h"
-#include "src/core/AccessWindowStatic.h"
-#include "src/core/helpers/WindowHelpers.h"
-
-#include <cstddef>
-#include <set>
-#include <sstream>
-#include <string>
-
-using namespace arm_compute;
-
-namespace
-{
-inline void options_add_matrix(std::set<std::string> &options, const std::array<float, 9> &matrix)
-{
-    for(size_t i = 0; i < 9; ++i)
-    {
-        std::stringstream mat_str;
-        mat_str << "-DMAT" << i << "=" << matrix[i] << " ";
-        options.insert(mat_str.str());
-    }
-}
-} // namespace
-
-BorderSize CLWarpPerspectiveKernel::border_size() const
-{
-    return BorderSize(1);
-}
-
-void CLWarpPerspectiveKernel::configure(const ICLTensor *input, ICLTensor *output, const std::array<float, 9> &matrix, InterpolationPolicy policy)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, matrix, policy);
-}
-
-void CLWarpPerspectiveKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const std::array<float, 9> &matrix, InterpolationPolicy policy)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON(InterpolationPolicy::AREA == policy);
-
-    _input  = input;
-    _output = output;
-
-    // Create build options
-    std::set<std::string> options;
-    options_add_matrix(options, matrix);
-    options.emplace(("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type())));
-
-    // Create kernel
-    std::string interpolation_name = string_from_interpolation_policy(policy);
-    std::transform(interpolation_name.begin(), interpolation_name.end(), interpolation_name.begin(), ::tolower);
-    std::string kernel_name = "warp_perspective_" + interpolation_name;
-    _kernel                 = create_kernel(compile_context, kernel_name, options);
-
-    // Set static kernel arguments
-    unsigned int idx = 2 * num_arguments_per_2D_tensor(); //Skip the input and output parameters
-    _kernel.setArg<cl_int>(idx++, input->info()->dimension(0));
-    _kernel.setArg<cl_int>(idx++, input->info()->dimension(1));
-
-    // Configure kernel window
-    constexpr unsigned int num_elems_processed_per_iteration = 4;
-
-    Window win = calculate_max_window(*output->info(), Steps(num_elems_processed_per_iteration));
-
-    AccessWindowStatic     input_access(input->info(), -border_size().left, -border_size().top, input->info()->dimension(0) + border_size().right, input->info()->dimension(1) + border_size().bottom);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    output_access.set_valid_region(win, ValidRegion(Coordinates(), output->info()->tensor_shape()));
-
-    ICLKernel::configure_internal(win);
-}
diff --git a/src/core/CL/kernels/CLWarpPerspectiveKernel.h b/src/core/CL/kernels/CLWarpPerspectiveKernel.h
deleted file mode 100644
index dcbe1c5560..0000000000
--- a/src/core/CL/kernels/CLWarpPerspectiveKernel.h
+++ /dev/null
@@ -1,59 +0,0 @@
-/*
- * Copyright (c) 2016-2020 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLWARPERSPECTIVEKERNEL_H
-#define ARM_COMPUTE_CLWARPERSPECTIVEKERNEL_H
-
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLSimple2DKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-/** Interface for the warp perspective kernel.*/
-class CLWarpPerspectiveKernel : public ICLSimple2DKernel
-{
-public:
-    /** Initialize the function's source, destination, interpolation policy and border_mode.
-     *
-     * @param[in]  input  Source tensor. Data types supported: U8.
-     * @param[out] output Destination tensor, Data types supported: U8.
-     * @param[in]  matrix The perspective matrix. Must be 3x3 of type float.
-     * @param[in]  policy The interpolation type.
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, const std::array<float, 9> &matrix, InterpolationPolicy policy);
-    /** Initialize the function's source, destination, interpolation policy and border_mode.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Source tensor. Data types supported: U8.
-     * @param[out] output          Destination tensor, Data types supported: U8.
-     * @param[in]  matrix          The perspective matrix. Must be 3x3 of type float.
-     * @param[in]  policy          The interpolation type.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const std::array<float, 9> &matrix, InterpolationPolicy policy);
-
-    // Inherited methods overridden:
-    BorderSize border_size() const override;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLWARPERSPECTIVEKERNEL_H */