COMPMID-1330: Add support for NormalizePlanarYUV operator in CL

Change-Id: Id0754b9e2bc3ef7ff2c4c21c3b89709588c41bd3
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/146637
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-by: Giorgio Arena <giorgio.arena@arm.com>

19 files changed, 860 insertions, 70 deletions

diff --git a/arm_compute/core/CL/CLKernels.h b/arm_compute/core/CL/CLKernels.h
index da2d3166a4..4750031603 100644
--- a/arm_compute/core/CL/CLKernels.h
+++ b/arm_compute/core/CL/CLKernels.h
@@ -96,6 +96,7 @@
 #include "arm_compute/core/CL/kernels/CLNonLinearFilterKernel.h"
 #include "arm_compute/core/CL/kernels/CLNonMaximaSuppression3x3Kernel.h"
 #include "arm_compute/core/CL/kernels/CLNormalizationLayerKernel.h"
+#include "arm_compute/core/CL/kernels/CLNormalizePlanarYUVLayerKernel.h"
 #include "arm_compute/core/CL/kernels/CLPermuteKernel.h"
 #include "arm_compute/core/CL/kernels/CLPixelWiseMultiplicationKernel.h"
 #include "arm_compute/core/CL/kernels/CLPoolingLayerKernel.h"
diff --git a/arm_compute/core/CL/kernels/CLNormalizePlanarYUVLayerKernel.h b/arm_compute/core/CL/kernels/CLNormalizePlanarYUVLayerKernel.h
new file mode 100644
index 0000000000..5418d31a0c
--- /dev/null
+++ b/arm_compute/core/CL/kernels/CLNormalizePlanarYUVLayerKernel.h
@@ -0,0 +1,83 @@
+/*
+ * Copyright (c) 2018 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_CLNORMALIZEPLANARYUVLAYERKERNEL_H__
+#define __ARM_COMPUTE_CLNORMALIZEPLANARYUVLAYERKERNEL_H__
+
+#include "arm_compute/core/CL/ICLKernel.h"
+
+namespace arm_compute
+{
+class ICLTensor;
+
+/** Interface for the NormalizePlanarYUV layer kernel. */
+class CLNormalizePlanarYUVLayerKernel : public ICLKernel
+{
+public:
+    /** Constructor */
+    CLNormalizePlanarYUVLayerKernel();
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLNormalizePlanarYUVLayerKernel(const CLNormalizePlanarYUVLayerKernel &) = delete;
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLNormalizePlanarYUVLayerKernel &operator=(const CLNormalizePlanarYUVLayerKernel &) = delete;
+    /** Default Move Constructor. */
+    CLNormalizePlanarYUVLayerKernel(CLNormalizePlanarYUVLayerKernel &&) = default;
+    /** Default move assignment operator */
+    CLNormalizePlanarYUVLayerKernel &operator=(CLNormalizePlanarYUVLayerKernel &&) = default;
+    /** Default destructor */
+    ~CLNormalizePlanarYUVLayerKernel() = default;
+
+    /** Set the input and output tensors.
+     *
+     * @param[in]  input  Source tensor. 3 lower dimensions represent a single input with dimensions [width, height, channels].
+     *                    Data types supported: F16/F32.
+     * @param[out] output Destination tensor. Data type supported: same as @p input
+     * @param[in]  mean   Mean values tensor. 1 dimension with size equal to the number of input channels. Data types supported: same as @p input
+     * @param[in]  std    Standard deviation values tensor. 1 dimension with size equal to the number of input channels.
+     *                    Data types supported: same as @p input
+     */
+    void configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *mean, const ICLTensor *std);
+    /** Static function to check if given info will lead to a valid configuration of @ref CLNormalizePlanarYUVLayerKernel
+     *
+     * @param[in]  input  Source tensor info. 3 lower dimensions represent a single input with dimensions [width, height, channels].
+     *                    Data types supported: F16/F32.
+     * @param[out] output Destination tensor info. Data type supported: same as @p input
+     * @param[in]  mean   Mean values tensor info. 1 dimension with size equal to the number of input channels. Data types supported: same as @p input
+     * @param[in]  std    Standard deviation values tensor info. 1 dimension with size equal to the number of input channels.
+     *                    Data types supported: same as @p input
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *mean, const ITensorInfo *std);
+
+    // Inherited methods overridden:
+    void run(const Window &window, cl::CommandQueue &queue) override;
+
+private:
+    const ICLTensor *_input;
+    ICLTensor       *_output;
+    const ICLTensor *_mean;
+    const ICLTensor *_std;
+};
+} // namespace arm_compute
+#endif /*__ARM_COMPUTE_CLNORMALIZEPLANARYUVLAYERKERNEL_H__ */
diff --git a/arm_compute/core/GLES_COMPUTE/kernels/GCNormalizePlanarYUVLayerKernel.h b/arm_compute/core/GLES_COMPUTE/kernels/GCNormalizePlanarYUVLayerKernel.h
index 0d785ca0d4..7ffe5b20df 100644
--- a/arm_compute/core/GLES_COMPUTE/kernels/GCNormalizePlanarYUVLayerKernel.h
+++ b/arm_compute/core/GLES_COMPUTE/kernels/GCNormalizePlanarYUVLayerKernel.h
@@ -50,14 +50,26 @@ public:
 
     /** Set the input and output tensors.
      *
-     * @param[in]  input  Source tensor. 3 lower dimensions represent a single input with dimensions [width, height, FM].
+     * @param[in]  input  Source tensor. 3 lower dimensions represent a single input with dimensions [width, height, channels].
      *                    Data types supported: F16.
-     * @param[out] output Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input
-     * @param[in]  mean   Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]  sd     Standard deviation values tensor. 1 dimension with size equal to the feature maps [FM].
-     *                     Data types supported: Same as @p input
+     * @param[out] output Destination tensor. Data type supported: same as @p input
+     * @param[in]  mean   Mean values tensor. 1 dimension with size equal to the number of input channels. Data types supported: same as @p input
+     * @param[in]  std    Standard deviation values tensor. 1 dimension with size equal to the feature maps [FM].
+     *                    Data types supported: same as @p input
      */
-    void configure(const IGCTensor *input, IGCTensor *output, const IGCTensor *mean, const IGCTensor *sd);
+    void configure(const IGCTensor *input, IGCTensor *output, const IGCTensor *mean, const IGCTensor *std);
+    /** Static function to check if given info will lead to a valid configuration of @ref GCNormalizePlanarYUVLayerKernel
+     *
+     * @param[in]  input  Source tensor info. 3 lower dimensions represent a single input with dimensions [width, height, channels].
+     *                    Data types supported: F16.
+     * @param[out] output Destination tensor info. Data type supported: same as @p input
+     * @param[in]  mean   Mean values tensor info. 1 dimension with size equal to the number of input channels. Data types supported: same as @p input
+     * @param[in]  std    Standard deviation values tensor info. 1 dimension with size equal to the number of input channels.
+     *                    Data types supported: same as @p input
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *mean, const ITensorInfo *std);
 
     // Inherited methods overridden:
     void run(const Window &window) override;
@@ -66,7 +78,7 @@ private:
     const IGCTensor *_input;
     IGCTensor       *_output;
     const IGCTensor *_mean;
-    const IGCTensor *_sd;
+    const IGCTensor *_std;
 };
 }
 #endif /*__ARM_COMPUTE_GCNORMALIZEPLANARYUVLAYERKERNEL_H__ */
diff --git a/arm_compute/runtime/CL/CLFunctions.h b/arm_compute/runtime/CL/CLFunctions.h
index d2bfdfd7cb..02a4dab6f1 100644
--- a/arm_compute/runtime/CL/CLFunctions.h
+++ b/arm_compute/runtime/CL/CLFunctions.h
@@ -93,6 +93,7 @@
 #include "arm_compute/runtime/CL/functions/CLNonLinearFilter.h"
 #include "arm_compute/runtime/CL/functions/CLNonMaximaSuppression3x3.h"
 #include "arm_compute/runtime/CL/functions/CLNormalizationLayer.h"
+#include "arm_compute/runtime/CL/functions/CLNormalizePlanarYUVLayer.h"
 #include "arm_compute/runtime/CL/functions/CLOpticalFlow.h"
 #include "arm_compute/runtime/CL/functions/CLPermute.h"
 #include "arm_compute/runtime/CL/functions/CLPhase.h"
diff --git a/arm_compute/runtime/CL/functions/CLNormalizePlanarYUVLayer.h b/arm_compute/runtime/CL/functions/CLNormalizePlanarYUVLayer.h
new file mode 100644
index 0000000000..85f7d93ddf
--- /dev/null
+++ b/arm_compute/runtime/CL/functions/CLNormalizePlanarYUVLayer.h
@@ -0,0 +1,75 @@
+/*
+ * Copyright (c) 2018 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_CLNORMALIZEPLANARYUVLAYER_H__
+#define __ARM_COMPUTE_CLNORMALIZEPLANARYUVLAYER_H__
+
+#include "arm_compute/runtime/IFunction.h"
+
+#include "arm_compute/core/CL/kernels/CLNormalizePlanarYUVLayerKernel.h"
+#include "arm_compute/core/Types.h"
+
+namespace arm_compute
+{
+class ICLTensor;
+
+/** Basic function to run @ref CLNormalizePlanarYUVLayerKernel
+ *
+ *  @note The function simulates a NormalizePlanarYUV layer.
+ */
+class CLNormalizePlanarYUVLayer : public IFunction
+{
+public:
+    /** Default constructor */
+    CLNormalizePlanarYUVLayer();
+    /** Set the input and output tensors.
+     *
+     * @param[in]  input  Source tensor. 3 lower dimensions represent a single input with dimensions [width, height, channels].
+     *                    Data types supported: F16/F32.
+     * @param[out] output Destinationfeature tensor. Data type supported: same as @p input
+     * @param[in]  mean   Mean values tensor. 1 dimension with size equal to the number of input channels. Data types supported: Same as @p input
+     * @param[in]  std    Standard deviation values tensor. 1 dimension with size equal to the number of input channels.
+     *                    Data types supported: Same as @p input
+     */
+    void configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *mean, const ICLTensor *std);
+    /** Static function to check if given info will lead to a valid configuration of @ref CLNormalizePlanarYUVLayer
+     *
+     * @param[in]  input  Source tensor info. 3 lower dimensions represent a single input with dimensions [width, height, FM].
+     *                    Data types supported: F16/F32.
+     * @param[out] output Destination tensor info. Data type supported: same as @p input
+     * @param[in]  mean   Mean values tensor info. 1 dimension with size equal to the number of input channels. Data types supported: Same as @p input
+     * @param[in]  std    Standard deviation values tensor info. 1 dimension with size equal to the number of input channels.
+     *                    Data types supported: Same as @p input
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *mean, const ITensorInfo *std);
+
+    // Inherited methods overridden:
+    void run() override;
+
+private:
+    CLNormalizePlanarYUVLayerKernel _norm_kernel; /**< NormalizePlanarYUV layer kernel to run */
+};
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_CLNORMALIZEPLANARYUVLAYER_H__ */
diff --git a/arm_compute/runtime/GLES_COMPUTE/functions/GCNormalizePlanarYUVLayer.h b/arm_compute/runtime/GLES_COMPUTE/functions/GCNormalizePlanarYUVLayer.h
index 2862eeb9cd..d6cf4d0803 100644
--- a/arm_compute/runtime/GLES_COMPUTE/functions/GCNormalizePlanarYUVLayer.h
+++ b/arm_compute/runtime/GLES_COMPUTE/functions/GCNormalizePlanarYUVLayer.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -44,14 +44,26 @@ public:
     GCNormalizePlanarYUVLayer();
     /** Set the input and output tensors.
      *
-     * @param[in]  input  Source tensor. 3 lower dimensions represent a single input with dimensions [width, height, FM].
+     * @param[in]  input  Source tensor. 3 lower dimensions represent a single input with dimensions [width, height, channels].
      *                    Data types supported: F16.
-     * @param[out] output Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input
-     * @param[in]  mean   Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]  sd     Standard deviation values tensor. 1 dimension with size equal to the feature maps [FM].
-     *                     Data types supported: Same as @p input
+     * @param[out] output Destination tensor. Data type supported: same as @p input
+     * @param[in]  mean   Mean values tensor. 1 dimension with size equal to the number of input channels. Data types supported: same as @p input
+     * @param[in]  std    Standard deviation values tensor. 1 dimension with size equal to the number of input channels.
+     *                    Data types supported: same as @p input
      */
-    void configure(const IGCTensor *input, IGCTensor *output, const IGCTensor *mean, const IGCTensor *sd);
+    void configure(const IGCTensor *input, IGCTensor *output, const IGCTensor *mean, const IGCTensor *std);
+    /** Static function to check if given info will lead to a valid configuration of @ref CLNormalizePlanarYUVLayer
+     *
+     * @param[in]  input  Source tensor info. 3 lower dimensions represent a single input with dimensions [width, height, channels].
+     *                    Data types supported: F16/F32.
+     * @param[out] output Destination tensor info. Data type supported: same as @p input
+     * @param[in]  mean   Mean values tensor info. 1 dimension with size equal to the number of input channels. Data types supported: same as @p input
+     * @param[in]  std    Standard deviation values tensor info. 1 dimension with size equal to the number of input channels.
+     *                    Data types supported: same as @p input
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *mean, const ITensorInfo *std);
 
     // Inherited methods overridden:
     void run() override;
diff --git a/src/core/CL/CLKernelLibrary.cpp b/src/core/CL/CLKernelLibrary.cpp
index 0cc6e320bf..4af2b09530 100644
--- a/src/core/CL/CLKernelLibrary.cpp
+++ b/src/core/CL/CLKernelLibrary.cpp
@@ -308,6 +308,8 @@ const std::map<std::string, std::string> CLKernelLibrary::_kernel_program_map =
     { "non_max_suppression", "nonmax.cl" },
     { "normalization_layer_cross_map", "normalization_layer.cl" },
     { "normalization_layer_in_map", "normalization_layer.cl" },
+    { "normalize_planar_yuv_layer_nchw", "normalize_planar_yuv_layer.cl" },
+    { "normalize_planar_yuv_layer_nhwc", "normalize_planar_yuv_layer.cl" },
     { "NV12_to_IYUV_bt709", "color_convert.cl" },
     { "NV12_to_RGB888_bt709", "color_convert.cl" },
     { "NV12_to_RGBA8888_bt709", "color_convert.cl" },
@@ -674,6 +676,10 @@ const std::map<std::string, std::string> CLKernelLibrary::_program_source_map =
 #include "./cl_kernels/normalization_layer.clembed"
     },
     {
+        "normalize_planar_yuv_layer.cl",
+#include "./cl_kernels/normalize_planar_yuv_layer.clembed"
+    },
+    {
         "batchnormalization_layer.cl",
 #include "./cl_kernels/batchnormalization_layer.clembed"
     },
diff --git a/src/core/CL/cl_kernels/normalize_planar_yuv_layer.cl b/src/core/CL/cl_kernels/normalize_planar_yuv_layer.cl
new file mode 100644
index 0000000000..dc6652449e
--- /dev/null
+++ b/src/core/CL/cl_kernels/normalize_planar_yuv_layer.cl
@@ -0,0 +1,134 @@
+/*
+ * Copyright (c) 2018 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 "helpers.h"
+
+#if defined(DATA_TYPE) && defined(VEC_SIZE)
+
+#define TYPE VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
+
+/** Apply normalize_planar_yuv layer on tensors with NCHW format.
+ *
+ * @note Data type should be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=float
+ * @note Vector size should be given as a preprocessor argument using -DVEC_SIZE e.g. -DVEC_SIZE=8
+ * @note The depth of the input tensor should be given as a preprocessor argument using -DNUM_CHANNELS e.g. -DNUM_CHANNELS=8
+ *
+ * @param[in]  src_ptr                            Pointer to the first source tensor. Supported data types: F16/F32
+ * @param[in]  src_stride_x                       Stride of the first source tensor in X dimension (in bytes)
+ * @param[in]  src_step_x                         input_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  src_stride_y                       Stride of the first source tensor in Y dimension (in bytes)
+ * @param[in]  src_step_y                         input_stride_y * number of elements along Y processed per workitem(in bytes)
+ * @param[in]  src_stride_z                       Stride of the first source tensor in Z dimension (in bytes)
+ * @param[in]  src_step_z                         input_stride_z * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  src_offset_first_element_in_bytes  The offset of the first element in the first source tensor
+ * @param[out] dst_ptr                            Pointer to the destination tensor. Supported data types: same as @p src_ptr
+ * @param[in]  dst_stride_x                       Stride of the destination tensor in X dimension (in bytes)
+ * @param[in]  dst_step_x                         output_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  dst_stride_y                       Stride of the destination tensor in Y dimension (in bytes)
+ * @param[in]  dst_step_y                         output_stride_y * number of elements along Y processed per workitem(in bytes)
+ * @param[in]  dst_stride_z                       Stride of the destination tensor in Z dimension (in bytes)
+ * @param[in]  dst_step_z                         output_stride_z * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  dst_offset_first_element_in_bytes  The offset of the first element in the destination tensor
+ * @param[in]  mean_ptr                           Pointer to the mean source tensor. Supported data types: same as @p src_ptr
+ * @param[in]  mean_stride_x                      Stride of the mean source tensor in X dimension (in bytes)
+ * @param[in]  mean_step_x                        mean_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  mean_offset_first_element_in_bytes The offset of the first element in the mean source tensor
+ * @param[in]  std_ptr                            Pointer to the std tensor. Supported data types: same as @p src_ptr
+ * @param[in]  std_stride_x                       Stride of the std tensor in X dimension (in bytes)
+ * @param[in]  std_step_x                         std_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  std_offset_first_element_in_bytes  The offset of the first element in the var source tensor
+ */
+__kernel void normalize_planar_yuv_layer_nchw(TENSOR3D_DECLARATION(src),
+                                              TENSOR3D_DECLARATION(dst),
+                                              VECTOR_DECLARATION(mean),
+                                              VECTOR_DECLARATION(std))
+{
+    Tensor3D src  = CONVERT_TO_TENSOR3D_STRUCT(src);
+    Tensor3D dst  = CONVERT_TO_TENSOR3D_STRUCT(dst);
+    Vector   mean = CONVERT_TO_VECTOR_STRUCT(mean);
+    Vector   std  = CONVERT_TO_VECTOR_STRUCT(std);
+
+    const uint current_slice = get_global_id(2) % NUM_CHANNELS;
+
+    const DATA_TYPE curr_mean = *((__global DATA_TYPE *)(mean.ptr + current_slice * mean.stride_x));
+    const DATA_TYPE curr_std  = *((__global DATA_TYPE *)(std.ptr + current_slice * std.stride_x));
+
+    TYPE data = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)src.ptr);
+    TYPE res  = (data - curr_mean) / curr_std;
+
+    VSTORE(VEC_SIZE)
+    (res, 0, (__global DATA_TYPE *)dst.ptr);
+}
+
+/** Apply normalize_planar_yuv layer on tensors with NHWC format.
+ *
+ * @note Data type should be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=float
+ * @note Vector size should be given as a preprocessor argument using -DVEC_SIZE e.g. -DVEC_SIZE=8
+ *
+ * @param[in]  src_ptr                            Pointer to the first source tensor. Supported data types: F16/F32
+ * @param[in]  src_stride_x                       Stride of the first source tensor in X dimension (in bytes)
+ * @param[in]  src_step_x                         input_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  src_stride_y                       Stride of the first source tensor in Y dimension (in bytes)
+ * @param[in]  src_step_y                         input_stride_y * number of elements along Y processed per workitem(in bytes)
+ * @param[in]  src_stride_z                       Stride of the first source tensor in Z dimension (in bytes)
+ * @param[in]  src_step_z                         input_stride_z * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  src_offset_first_element_in_bytes  The offset of the first element in the first source tensor
+ * @param[out] dst_ptr                            Pointer to the destination tensor. Supported data types: same as @p src_ptr
+ * @param[in]  dst_stride_x                       Stride of the destination tensor in X dimension (in bytes)
+ * @param[in]  dst_step_x                         output_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  dst_stride_y                       Stride of the destination tensor in Y dimension (in bytes)
+ * @param[in]  dst_step_y                         output_stride_y * number of elements along Y processed per workitem(in bytes)
+ * @param[in]  dst_stride_z                       Stride of the destination tensor in Z dimension (in bytes)
+ * @param[in]  dst_step_z                         output_stride_z * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  dst_offset_first_element_in_bytes  The offset of the first element in the destination tensor
+ * @param[in]  mean_ptr                           Pointer to the mean source tensor. Supported data types: same as @p src_ptr
+ * @param[in]  mean_stride_x                      Stride of the mean source tensor in X dimension (in bytes)
+ * @param[in]  mean_step_x                        mean_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  mean_offset_first_element_in_bytes The offset of the first element in the mean source tensor
+ * @param[in]  std_ptr                            Pointer to the std tensor. Supported data types: same as @p src_ptr
+ * @param[in]  std_stride_x                       Stride of the std tensor in X dimension (in bytes)
+ * @param[in]  std_step_x                         std_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  std_offset_first_element_in_bytes  The offset of the first element in the var source tensor
+ */
+__kernel void normalize_planar_yuv_layer_nhwc(TENSOR3D_DECLARATION(src),
+                                              TENSOR3D_DECLARATION(dst),
+                                              VECTOR_DECLARATION(mean),
+                                              VECTOR_DECLARATION(std))
+{
+    Tensor3D src  = CONVERT_TO_TENSOR3D_STRUCT(src);
+    Tensor3D dst  = CONVERT_TO_TENSOR3D_STRUCT(dst);
+    Vector   mean = CONVERT_TO_VECTOR_STRUCT(mean);
+    Vector   std  = CONVERT_TO_VECTOR_STRUCT(std);
+
+    const uint current_slice = get_global_id(0);
+
+    const TYPE curr_mean = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)(mean.ptr + current_slice * VEC_SIZE * mean.stride_x));
+    const TYPE curr_std  = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)(std.ptr + current_slice * VEC_SIZE * std.stride_x));
+
+    TYPE data = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)src.ptr);
+    TYPE res  = (data - curr_mean) / curr_std;
+
+    VSTORE(VEC_SIZE)
+    (res, 0, (__global DATA_TYPE *)dst.ptr);
+}
+#endif // defined(DATA_TYPE) && defined(VEC_SIZE)
diff --git a/src/core/CL/kernels/CLNormalizePlanarYUVLayerKernel.cpp b/src/core/CL/kernels/CLNormalizePlanarYUVLayerKernel.cpp
new file mode 100644
index 0000000000..31451ef422
--- /dev/null
+++ b/src/core/CL/kernels/CLNormalizePlanarYUVLayerKernel.cpp
@@ -0,0 +1,173 @@
+/*
+ * Copyright (c) 2018 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/kernels/CLNormalizePlanarYUVLayerKernel.h"
+
+#include "arm_compute/core/AccessWindowStatic.h"
+#include "arm_compute/core/CL/CLHelpers.h"
+#include "arm_compute/core/CL/CLKernelLibrary.h"
+#include "arm_compute/core/CL/CLValidate.h"
+#include "arm_compute/core/CL/ICLTensor.h"
+#include "arm_compute/core/Helpers.h"
+#include "arm_compute/core/TensorInfo.h"
+#include "arm_compute/core/Utils.h"
+#include "arm_compute/core/Window.h"
+
+#include "support/ToolchainSupport.h"
+
+using namespace arm_compute;
+
+namespace
+{
+Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *mean, const ITensorInfo *std)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(output);
+
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, mean, std);
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(mean, std);
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(mean->num_dimensions() > 1, "mean and std must be vectors");
+
+    const unsigned int channel_idx = get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::CHANNEL);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(channel_idx) != mean->dimension(0));
+
+    // Checks performed when output is configured
+    if(output->total_size() != 0)
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input, output);
+    }
+
+    return Status{};
+}
+
+std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output, ITensorInfo *mean, ITensorInfo *std)
+{
+    // Output tensor auto initialization if not yet initialized
+    auto_init_if_empty(*output, *input->clone());
+
+    const unsigned int num_elems_processed_per_iteration = 16 / input->element_size();
+
+    Window win = calculate_max_window(*input, Steps(num_elems_processed_per_iteration));
+
+    AccessWindowHorizontal input_access(input, 0, num_elems_processed_per_iteration);
+    AccessWindowHorizontal output_access(output, 0, num_elems_processed_per_iteration);
+
+    bool window_changed = update_window_and_padding(win, input_access, output_access);
+    output_access.set_valid_region(win, input->valid_region());
+
+    if(input->data_layout() == DataLayout::NHWC)
+    {
+        AccessWindowHorizontal mean_access(mean, 0, num_elems_processed_per_iteration);
+        AccessWindowHorizontal std_access(std, 0, num_elems_processed_per_iteration);
+        window_changed = window_changed || update_window_and_padding(win, mean_access, std_access);
+    }
+
+    Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
+    return std::make_pair(err, win);
+}
+} // namespace
+
+CLNormalizePlanarYUVLayerKernel::CLNormalizePlanarYUVLayerKernel()
+    : _input(nullptr), _output(nullptr), _mean(nullptr), _std(nullptr)
+{
+}
+
+void CLNormalizePlanarYUVLayerKernel::configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *mean, const ICLTensor *std)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output, mean, std);
+
+    // Output tensor auto initialization if not yet initialized
+    auto_init_if_empty(*output->info(), *input->info()->clone());
+
+    // Perform validation step
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), mean->info(), std->info()));
+
+    _input  = input;
+    _output = output;
+    _mean   = mean;
+    _std    = std;
+
+    const unsigned int num_elems_processed_per_iteration = 16 / input->info()->element_size();
+    const unsigned int channel_idx                       = get_data_layout_dimension_index(input->info()->data_layout(), DataLayoutDimension::CHANNEL);
+
+    // Set build options
+    CLBuildOptions build_opts;
+    build_opts.add_option(("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type())));
+    build_opts.add_option(("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration)));
+    build_opts.add_option(("-DNUM_CHANNELS=" + support::cpp11::to_string(input->info()->dimension(channel_idx))));
+
+    // Create kernel
+    _kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel("normalize_planar_yuv_layer_" + lower_string(string_from_data_layout(input->info()->data_layout())), build_opts.options()));
+
+    // Configure kernel window
+    auto win_config = validate_and_configure_window(input->info(), output->info(), mean->info(), std->info());
+    ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
+    ICLKernel::configure_internal(win_config.second);
+
+    // Set config_id for enabling LWS tuning
+    _config_id = "normalize_planar_yuv_layer_";
+    _config_id += lower_string(string_from_data_layout(input->info()->data_layout()));
+    _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));
+}
+
+Status CLNormalizePlanarYUVLayerKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *mean, const ITensorInfo *std)
+{
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, mean, std));
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get(), mean->clone().get(), std->clone().get()).first);
+
+    return Status{};
+}
+
+void CLNormalizePlanarYUVLayerKernel::run(const Window &window, cl::CommandQueue &queue)
+{
+    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
+    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
+
+    Window collapsed = window.collapse_if_possible(ICLKernel::window(), Window::DimZ);
+    Window slice     = collapsed.first_slice_window_3D();
+
+    Window slice_in = collapsed.first_slice_window_1D();
+    slice_in.set(Window::DimX, Window::Dimension(0, 0, 0));
+
+    unsigned int idx = 2 * num_arguments_per_3D_tensor();
+    add_1D_tensor_argument(idx, _mean, slice_in);
+    add_1D_tensor_argument(idx, _std, slice_in);
+
+    do
+    {
+        idx = 0;
+        add_3D_tensor_argument(idx, _input, slice);
+        add_3D_tensor_argument(idx, _output, slice);
+        enqueue(queue, *this, slice, lws_hint());
+    }
+    while(collapsed.slide_window_slice_3D(slice));
+}
diff --git a/src/core/GLES_COMPUTE/kernels/GCNormalizePlanarYUVLayerKernel.cpp b/src/core/GLES_COMPUTE/kernels/GCNormalizePlanarYUVLayerKernel.cpp
index fac29024e3..03463b2552 100644
--- a/src/core/GLES_COMPUTE/kernels/GCNormalizePlanarYUVLayerKernel.cpp
+++ b/src/core/GLES_COMPUTE/kernels/GCNormalizePlanarYUVLayerKernel.cpp
@@ -36,26 +36,75 @@
 
 using namespace arm_compute;
 
+namespace
+{
+Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *mean, const ITensorInfo *std)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F16);
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(output);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_layout() != DataLayout::NCHW);
+
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, mean, std);
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(mean, std);
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(mean->num_dimensions() > 1, "mean and std must be vectors");
+
+    const unsigned int channel_idx = get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::CHANNEL);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(channel_idx) != mean->dimension(0));
+
+    // Checks performed when output is configured
+    if(output->total_size() != 0)
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input, output);
+    }
+
+    return Status{};
+}
+
+std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output, ITensorInfo *mean, ITensorInfo *std)
+{
+    // Output tensor auto initialization if not yet initialized
+    auto_init_if_empty(*output, *input->clone());
+
+    const unsigned int num_elems_processed_per_iteration = 4;
+
+    // Configure kernel window
+    Window win = calculate_max_window(*input, Steps(num_elems_processed_per_iteration));
+
+    AccessWindowHorizontal input_access(input, 0, num_elems_processed_per_iteration);
+    AccessWindowHorizontal output_access(output, 0, num_elems_processed_per_iteration);
+    const int              mean_padding = ceil_to_multiple(mean->dimension(0), num_elems_processed_per_iteration) - mean->dimension(0);
+    const int              std_padding  = ceil_to_multiple(std->dimension(0), num_elems_processed_per_iteration) - std->dimension(0);
+    AccessWindowStatic     mean_access(mean, 0, 0, mean->dimension(0) + mean_padding, mean->dimension(1));
+    AccessWindowStatic     std_access(std, 0, 0, std->dimension(0) + std_padding, std->dimension(1));
+
+    const bool window_changed = update_window_and_padding(win, input_access, output_access, mean_access, std_access);
+    output_access.set_valid_region(win, input->valid_region());
+
+    Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
+    return std::make_pair(err, win);
+}
+} // namespace
+
 GCNormalizePlanarYUVLayerKernel::GCNormalizePlanarYUVLayerKernel()
-    : _input(nullptr), _output(nullptr), _mean(nullptr), _sd(nullptr)
+    : _input(nullptr), _output(nullptr), _mean(nullptr), _std(nullptr)
 {
 }
 
-void GCNormalizePlanarYUVLayerKernel::configure(const IGCTensor *input, IGCTensor *output, const IGCTensor *mean, const IGCTensor *sd)
+void GCNormalizePlanarYUVLayerKernel::configure(const IGCTensor *input, IGCTensor *output, const IGCTensor *mean, const IGCTensor *std)
 {
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F16);
-    ARM_COMPUTE_ERROR_ON_NULLPTR(output);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output, mean, sd);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_SHAPES(input, output);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_SHAPES(mean, sd);
-    ARM_COMPUTE_ERROR_ON(input->info()->dimension(2) != mean->info()->dimension(0));
+    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output, mean, std);
+
+    // Output tensor auto initialization if not yet initialized
+    auto_init_if_empty(*output->info(), *input->info()->clone());
+
+    // Perform validation step
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), mean->info(), std->info()));
 
     _input  = input;
     _output = output;
     _mean   = mean;
-    _sd     = sd;
-
-    const unsigned int num_elems_processed_per_iteration = 4;
+    _std    = std;
 
     // Set build options
     std::set<std::string> build_opts;
@@ -67,19 +116,17 @@ void GCNormalizePlanarYUVLayerKernel::configure(const IGCTensor *input, IGCTenso
     _kernel = static_cast<GCKernel>(GCKernelLibrary::get().create_kernel("normalize_planar_yuv_layer", build_opts));
 
     // Configure kernel window
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration));
+    auto win_config = validate_and_configure_window(input->info(), output->info(), mean->info(), std->info());
+    ARM_COMPUTE_ERROR_THROW_ON(std::get<0>(win_config));
 
-    AccessWindowHorizontal input_access(input->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
-    const int              mean_padding = ceil_to_multiple(mean->info()->dimension(0), num_elems_processed_per_iteration) - mean->info()->dimension(0);
-    const int              sd_padding   = ceil_to_multiple(sd->info()->dimension(0), num_elems_processed_per_iteration) - sd->info()->dimension(0);
-    AccessWindowStatic     mean_access(mean->info(), 0, 0, mean->info()->dimension(0) + mean_padding, mean->info()->dimension(1));
-    AccessWindowStatic     sd_access(sd->info(), 0, 0, sd->info()->dimension(0) + sd_padding, sd->info()->dimension(1));
-
-    update_window_and_padding(win, input_access, output_access, mean_access, sd_access);
-    output_access.set_valid_region(win, input->info()->valid_region());
+    IGCKernel::configure(std::get<1>(win_config));
+}
 
-    IGCKernel::configure(win);
+Status GCNormalizePlanarYUVLayerKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *mean, const ITensorInfo *std)
+{
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, mean, std));
+    ARM_COMPUTE_RETURN_ON_ERROR(std::get<0>(validate_and_configure_window(input->clone().get(), output->clone().get(), mean->clone().get(), std->clone().get())));
+    return Status{};
 }
 
 void GCNormalizePlanarYUVLayerKernel::run(const Window &window)
@@ -100,7 +147,7 @@ void GCNormalizePlanarYUVLayerKernel::run(const Window &window)
 
     unsigned int idx = 2 * num_arguments_per_3D_tensor();
     add_1D_tensor_argument(idx, _mean, 3, slice_in);
-    add_1D_tensor_argument(idx, _sd, 4, slice_in);
+    add_1D_tensor_argument(idx, _std, 4, slice_in);
 
     slice_in = window.first_slice_window_3D();
 
diff --git a/src/runtime/CL/functions/CLNormalizePlanarYUVLayer.cpp b/src/runtime/CL/functions/CLNormalizePlanarYUVLayer.cpp
new file mode 100644
index 0000000000..11d70e31fb
--- /dev/null
+++ b/src/runtime/CL/functions/CLNormalizePlanarYUVLayer.cpp
@@ -0,0 +1,55 @@
+/*
+ * Copyright (c) 2018 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/runtime/CL/functions/CLNormalizePlanarYUVLayer.h"
+
+#include "arm_compute/core/Error.h"
+#include "arm_compute/core/TensorInfo.h"
+#include "arm_compute/core/Types.h"
+#include "arm_compute/core/Validate.h"
+#include "arm_compute/runtime/CL/CLScheduler.h"
+
+namespace arm_compute
+{
+CLNormalizePlanarYUVLayer::CLNormalizePlanarYUVLayer()
+    : _norm_kernel()
+{
+}
+
+void CLNormalizePlanarYUVLayer::configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *mean, const ICLTensor *std)
+{
+    _norm_kernel.configure(input, output, mean, std);
+}
+
+Status CLNormalizePlanarYUVLayer::validate(const ITensorInfo *input, const ITensorInfo *output,
+                                           const ITensorInfo *mean, const ITensorInfo *std)
+{
+    return CLNormalizePlanarYUVLayerKernel::validate(input, output, mean, std);
+}
+
+void CLNormalizePlanarYUVLayer::run()
+{
+    CLScheduler::get().enqueue(_norm_kernel, true);
+}
+} // namespace arm_compute
diff --git a/src/runtime/GLES_COMPUTE/functions/GCNormalizePlanarYUVLayer.cpp b/src/runtime/GLES_COMPUTE/functions/GCNormalizePlanarYUVLayer.cpp
index 5fb971c154..19fdc3d7c0 100755
--- a/src/runtime/GLES_COMPUTE/functions/GCNormalizePlanarYUVLayer.cpp
+++ b/src/runtime/GLES_COMPUTE/functions/GCNormalizePlanarYUVLayer.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -37,9 +37,15 @@ GCNormalizePlanarYUVLayer::GCNormalizePlanarYUVLayer()
 {
 }
 
-void GCNormalizePlanarYUVLayer::configure(const IGCTensor *input, IGCTensor *output, const IGCTensor *mean, const IGCTensor *sd)
+void GCNormalizePlanarYUVLayer::configure(const IGCTensor *input, IGCTensor *output, const IGCTensor *mean, const IGCTensor *std)
 {
-    _norm_kernel.configure(input, output, mean, sd);
+    _norm_kernel.configure(input, output, mean, std);
+}
+
+Status GCNormalizePlanarYUVLayer::validate(const ITensorInfo *input, const ITensorInfo *output,
+                                           const ITensorInfo *mean, const ITensorInfo *std)
+{
+    return GCNormalizePlanarYUVLayerKernel::validate(input, output, mean, std);
 }
 
 void GCNormalizePlanarYUVLayer::run()
diff --git a/tests/datasets/NormalizePlanarYUVLayerDataset.h b/tests/datasets/NormalizePlanarYUVLayerDataset.h
index 2d71a56a30..1a97e68a92 100644
--- a/tests/datasets/NormalizePlanarYUVLayerDataset.h
+++ b/tests/datasets/NormalizePlanarYUVLayerDataset.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -55,7 +55,7 @@ public:
             description << "In=" << *_tensor_it << ":";
             description << "Out=" << *_tensor_it << ":";
             description << "Mean=" << *_param_it << ":";
-            description << "Sd=" << *_param_it << ":";
+            description << "Std=" << *_param_it << ":";
             return description.str();
         }
 
diff --git a/tests/datasets/RandomNormalizePlanarYUVLayerDataset.h b/tests/datasets/RandomNormalizePlanarYUVLayerDataset.h
index 5693004070..56eb604cca 100644
--- a/tests/datasets/RandomNormalizePlanarYUVLayerDataset.h
+++ b/tests/datasets/RandomNormalizePlanarYUVLayerDataset.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -46,6 +46,8 @@ public:
         add_config(TensorShape(21U, 11U, 12U, 1U), TensorShape(12U));
         add_config(TensorShape(7U, 3U, 6U, 1U), TensorShape(6U));
         add_config(TensorShape(7U, 2U, 3U, 1U), TensorShape(3U));
+        add_config(TensorShape(7U, 2U, 3U, 3U), TensorShape(3U));
+        add_config(TensorShape(21U, 11U, 12U, 3U), TensorShape(12U));
     }
 };
 } // namespace datasets
diff --git a/tests/validation/CL/NormalizePlanarYUVLayer.cpp b/tests/validation/CL/NormalizePlanarYUVLayer.cpp
new file mode 100644
index 0000000000..aa1a00e106
--- /dev/null
+++ b/tests/validation/CL/NormalizePlanarYUVLayer.cpp
@@ -0,0 +1,142 @@
+/*
+ * Copyright (c) 2018 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/Types.h"
+#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/CLTensorAllocator.h"
+#include "arm_compute/runtime/CL/functions/CLNormalizePlanarYUVLayer.h"
+#include "tests/CL/CLAccessor.h"
+#include "tests/PaddingCalculator.h"
+#include "tests/datasets/RandomNormalizePlanarYUVLayerDataset.h"
+#include "tests/datasets/ShapeDatasets.h"
+#include "tests/framework/Asserts.h"
+#include "tests/framework/Macros.h"
+#include "tests/framework/datasets/Datasets.h"
+#include "tests/validation/Validation.h"
+#include "tests/validation/fixtures/NormalizePlanarYUVLayerFixture.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace
+{
+constexpr RelativeTolerance<float> tolerance_f16(0.001f); /**< Tolerance value for comparing reference's output against implementation's output for DataType::F16 */
+constexpr RelativeTolerance<float> tolerance_f32(0.001f); /**< Tolerance value for comparing reference's output against implementation's output for DataType::F32 */
+} // namespace
+
+TEST_SUITE(CL)
+TEST_SUITE(NormalizePlanarYUVLayer)
+
+template <typename T>
+using CLNormalizePlanarYUVLayerFixture = NormalizePlanarYUVLayerValidationFixture<CLTensor, CLAccessor, CLNormalizePlanarYUVLayer, T>;
+
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::RandomNormalizePlanarYUVLayerDataset(), framework::dataset::make("DataType", { DataType::F16 })),
+                                                                   framework::dataset::make("DataLayout", { DataLayout::NCHW, DataLayout::NHWC })),
+               shape0, shape1, dt, data_layout)
+{
+    TensorShape src_dst_shapes = shape0;
+    if(data_layout == DataLayout::NHWC)
+    {
+        permute(src_dst_shapes, PermutationVector(2U, 0U, 1U));
+    }
+
+    // Create tensors
+    CLTensor src  = create_tensor<CLTensor>(src_dst_shapes, dt, 1, QuantizationInfo(), data_layout);
+    CLTensor dst  = create_tensor<CLTensor>(src_dst_shapes, dt, 1, QuantizationInfo(), data_layout);
+    CLTensor mean = create_tensor<CLTensor>(shape1, dt, 1);
+    CLTensor sd   = create_tensor<CLTensor>(shape1, dt, 1);
+
+    // Create and Configure function
+    CLNormalizePlanarYUVLayer norm;
+    norm.configure(&src, &dst, &mean, &sd);
+
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(src_dst_shapes);
+    validate(dst.info()->valid_region(), valid_region);
+}
+
+// *INDENT-OFF*
+// clang-format off
+DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
+                    framework::dataset::make("InputInfo", { TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F32),     // Mismatching data types
+                        TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F16),     // Window shrink
+                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),      // Unsupported data type
+                        TensorInfo(TensorShape(32U, 16U, 8U), 1, DataType::F16),
+                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F16),     // Mismatching mean and sd shapes
+                        TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F32),     // Mismatching shapes
+                        }),
+                    framework::dataset::make("OutputInfo",{ TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F16),
+                        TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F16),
+                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                        TensorInfo(TensorShape(32U, 16U, 8U), 1, DataType::F16),
+                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F16),
+                        TensorInfo(TensorShape(30U, 11U, 2U), 1, DataType::F32),
+                        })),
+                framework::dataset::make("MSTDInfo",{ TensorInfo(TensorShape(2U), 1, DataType::F16),
+                    TensorInfo(TensorShape(2U), 1, DataType::F16),
+                    TensorInfo(TensorShape(2U), 1, DataType::U8),
+                    TensorInfo(TensorShape(8U), 1, DataType::F16),
+                    TensorInfo(TensorShape(6U), 1, DataType::F16),
+                    TensorInfo(TensorShape(2U), 1, DataType::F32),
+                    })),
+                    framework::dataset::make("Expected", { false, false, false, true, false, false })),
+                    input_info, output_info, msd_info, expected)
+{
+    const auto &mean_info = msd_info;
+    const auto &sd_info   = msd_info;
+    bool has_error = bool(CLNormalizePlanarYUVLayer::validate(&input_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false), &mean_info.clone()->set_is_resizable(false), &sd_info.clone()->set_is_resizable(false)));
+    ARM_COMPUTE_EXPECT(has_error == expected, framework::LogLevel::ERRORS);
+}
+// clang-format on
+// *INDENT-ON*
+
+TEST_SUITE(Float)
+TEST_SUITE(FP16)
+FIXTURE_DATA_TEST_CASE(Random, CLNormalizePlanarYUVLayerFixture<half>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::RandomNormalizePlanarYUVLayerDataset(),
+                                                                                                                  framework::dataset::make("DataType", DataType::F16)),
+                                                                                                                  framework::dataset::make("DataLayout", { DataLayout::NCHW, DataLayout::NHWC })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_f16, 0);
+}
+TEST_SUITE_END()
+
+TEST_SUITE(FP32)
+FIXTURE_DATA_TEST_CASE(Random, CLNormalizePlanarYUVLayerFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::RandomNormalizePlanarYUVLayerDataset(),
+                                                                                                                   framework::dataset::make("DataType", DataType::F32)),
+                                                                                                                   framework::dataset::make("DataLayout", { DataLayout::NCHW, DataLayout::NHWC })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_f32);
+}
+TEST_SUITE_END()
+TEST_SUITE_END()
+
+TEST_SUITE_END()
+TEST_SUITE_END()
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/GLES_COMPUTE/NormalizePlanarYUVLayer.cpp b/tests/validation/GLES_COMPUTE/NormalizePlanarYUVLayer.cpp
index e06b19cfea..540a2be143 100644
--- a/tests/validation/GLES_COMPUTE/NormalizePlanarYUVLayer.cpp
+++ b/tests/validation/GLES_COMPUTE/NormalizePlanarYUVLayer.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -70,10 +70,46 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::Ran
     validate(dst.info()->valid_region(), valid_region);
 }
 
+// *INDENT-OFF*
+// clang-format off
+DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
+                    framework::dataset::make("InputInfo", { TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F16),     // Mismatching data types
+                        TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F16),     // Window shrink
+                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),      // Unsupported data type
+                        TensorInfo(TensorShape(32U, 16U, 8U), 1, DataType::F16),
+                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F16),     // Mismatching mean and sd shapes
+                        TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F16),     // Mismatching shapes
+                        }),
+                    framework::dataset::make("OutputInfo",{ TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F32),
+                        TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F16),
+                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                        TensorInfo(TensorShape(32U, 16U, 8U), 1, DataType::F16),
+                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F16),
+                        TensorInfo(TensorShape(30U, 11U, 2U), 1, DataType::F16),
+                        })),
+                framework::dataset::make("MSTDInfo",{ TensorInfo(TensorShape(2U), 1, DataType::F16),
+                    TensorInfo(TensorShape(2U), 1, DataType::F16),
+                    TensorInfo(TensorShape(2U), 1, DataType::U8),
+                    TensorInfo(TensorShape(8U), 1, DataType::F16),
+                    TensorInfo(TensorShape(6U), 1, DataType::F16),
+                    TensorInfo(TensorShape(2U), 1, DataType::F16),
+                    })),
+                    framework::dataset::make("Expected", { false, false, false, true, false, false })),
+                    input_info, output_info, msd_info, expected)
+{
+    const auto &mean_info = msd_info;
+    const auto &sd_info   = msd_info;
+    bool has_error = bool(GCNormalizePlanarYUVLayer::validate(&input_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false), &mean_info.clone()->set_is_resizable(false), &sd_info.clone()->set_is_resizable(false)));
+    ARM_COMPUTE_EXPECT(has_error == expected, framework::LogLevel::ERRORS);
+}
+// clang-format on
+// *INDENT-ON*
+
 TEST_SUITE(Float)
 TEST_SUITE(FP16)
-FIXTURE_DATA_TEST_CASE(Random, GCNormalizePlanarYUVLayerFixture<half>, framework::DatasetMode::PRECOMMIT, combine(datasets::RandomNormalizePlanarYUVLayerDataset(),
-                                                                                                                  framework::dataset::make("DataType", DataType::F16)))
+FIXTURE_DATA_TEST_CASE(Random, GCNormalizePlanarYUVLayerFixture<half>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::RandomNormalizePlanarYUVLayerDataset(),
+                                                                                                                  framework::dataset::make("DataType", DataType::F16)),
+                                                                                                                  framework::dataset::make("DataLayout", { DataLayout::NCHW })))
 {
     // Validate output
     validate(GCAccessor(_target), _reference, tolerance_f16, 0);
diff --git a/tests/validation/fixtures/NormalizePlanarYUVLayerFixture.h b/tests/validation/fixtures/NormalizePlanarYUVLayerFixture.h
index 09905cfef7..cc73e530ef 100644
--- a/tests/validation/fixtures/NormalizePlanarYUVLayerFixture.h
+++ b/tests/validation/fixtures/NormalizePlanarYUVLayerFixture.h
@@ -45,16 +45,16 @@ class NormalizePlanarYUVLayerValidationFixture : public framework::Fixture
 {
 public:
     template <typename...>
-    void setup(TensorShape shape0, TensorShape shape1, DataType dt)
+    void setup(TensorShape shape0, TensorShape shape1, DataType dt, DataLayout data_layout)
     {
         _data_type = dt;
-        _target    = compute_target(shape0, shape1, dt);
+        _target    = compute_target(shape0, shape1, dt, data_layout);
         _reference = compute_reference(shape0, shape1, dt);
     }
 
 protected:
     template <typename U>
-    void fill(U &&src_tensor, U &&mean_tensor, U &&sd_tensor)
+    void fill(U &&src_tensor, U &&mean_tensor, U &&std_tensor)
     {
         if(is_data_type_float(_data_type))
         {
@@ -62,43 +62,48 @@ protected:
             float max_bound = 0.f;
             std::tie(min_bound, max_bound) = get_normalize_planar_yuv_layer_test_bounds<T>();
             std::uniform_real_distribution<> distribution(min_bound, max_bound);
-            std::uniform_real_distribution<> distribution_sd(0.1, max_bound);
+            std::uniform_real_distribution<> distribution_std(0.1, max_bound);
             library->fill(src_tensor, distribution, 0);
             library->fill(mean_tensor, distribution, 1);
-            library->fill(sd_tensor, distribution_sd, 2);
+            library->fill(std_tensor, distribution_std, 2);
         }
     }
 
-    TensorType compute_target(const TensorShape &shape0, const TensorShape &shape1, DataType dt)
+    TensorType compute_target(TensorShape shape0, const TensorShape &shape1, DataType dt, DataLayout data_layout)
     {
+        if(data_layout == DataLayout::NHWC)
+        {
+            permute(shape0, PermutationVector(2U, 0U, 1U));
+        }
+
         // Create tensors
-        TensorType src  = create_tensor<TensorType>(shape0, dt, 1);
-        TensorType dst  = create_tensor<TensorType>(shape0, dt, 1);
+        TensorType src  = create_tensor<TensorType>(shape0, dt, 1, QuantizationInfo(), data_layout);
+        TensorType dst  = create_tensor<TensorType>(shape0, dt, 1, QuantizationInfo(), data_layout);
         TensorType mean = create_tensor<TensorType>(shape1, dt, 1);
-        TensorType sd   = create_tensor<TensorType>(shape1, dt, 1);
+        TensorType std  = create_tensor<TensorType>(shape1, dt, 1);
 
         // Create and configure function
         FunctionType norm;
-        norm.configure(&src, &dst, &mean, &sd);
+        norm.configure(&src, &dst, &mean, &std);
 
         ARM_COMPUTE_EXPECT(src.info()->is_resizable(), framework::LogLevel::ERRORS);
         ARM_COMPUTE_EXPECT(dst.info()->is_resizable(), framework::LogLevel::ERRORS);
         ARM_COMPUTE_EXPECT(mean.info()->is_resizable(), framework::LogLevel::ERRORS);
-        ARM_COMPUTE_EXPECT(sd.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(std.info()->is_resizable(), framework::LogLevel::ERRORS);
 
         // Allocate tensors
         src.allocator()->allocate();
         dst.allocator()->allocate();
         mean.allocator()->allocate();
-        sd.allocator()->allocate();
+        std.allocator()->allocate();
 
         ARM_COMPUTE_EXPECT(!src.info()->is_resizable(), framework::LogLevel::ERRORS);
         ARM_COMPUTE_EXPECT(!dst.info()->is_resizable(), framework::LogLevel::ERRORS);
         ARM_COMPUTE_EXPECT(!mean.info()->is_resizable(), framework::LogLevel::ERRORS);
-        ARM_COMPUTE_EXPECT(!sd.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!std.info()->is_resizable(), framework::LogLevel::ERRORS);
 
         // Fill tensors
-        fill(AccessorType(src), AccessorType(mean), AccessorType(sd));
+        fill(AccessorType(src), AccessorType(mean), AccessorType(std));
 
         // Compute function
         norm.run();
@@ -111,12 +116,12 @@ protected:
         // Create reference
         SimpleTensor<T> ref_src{ shape0, dt, 1 };
         SimpleTensor<T> ref_mean{ shape1, dt, 1 };
-        SimpleTensor<T> ref_sd{ shape1, dt, 1 };
+        SimpleTensor<T> ref_std{ shape1, dt, 1 };
 
         // Fill reference
-        fill(ref_src, ref_mean, ref_sd);
+        fill(ref_src, ref_mean, ref_std);
 
-        return reference::normalize_planar_yuv_layer(ref_src, ref_mean, ref_sd);
+        return reference::normalize_planar_yuv_layer(ref_src, ref_mean, ref_std);
     }
 
     TensorType      _target{};
diff --git a/tests/validation/reference/NormalizePlanarYUVLayer.cpp b/tests/validation/reference/NormalizePlanarYUVLayer.cpp
index 2442943bb4..afb899220d 100644
--- a/tests/validation/reference/NormalizePlanarYUVLayer.cpp
+++ b/tests/validation/reference/NormalizePlanarYUVLayer.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -35,7 +35,7 @@ namespace reference
 {
 // NormalizePlanarYUV Layer for floating point type
 template <typename T, typename std::enable_if<is_floating_point<T>::value, int>::type *>
-SimpleTensor<T> normalize_planar_yuv_layer(const SimpleTensor<T> &src, const SimpleTensor<T> &mean, const SimpleTensor<T> &sd)
+SimpleTensor<T> normalize_planar_yuv_layer(const SimpleTensor<T> &src, const SimpleTensor<T> &mean, const SimpleTensor<T> &std)
 {
     SimpleTensor<T> result(src.shape(), src.data_type());
 
@@ -53,7 +53,7 @@ SimpleTensor<T> normalize_planar_yuv_layer(const SimpleTensor<T> &src, const Sim
                 for(int l = 0; l < cols; ++l)
                 {
                     const int pos = l + k * cols + i * rows * cols + r * cols * rows * depth;
-                    result[pos]   = (src[pos] - mean[i]) / sd[i];
+                    result[pos]   = (src[pos] - mean[i]) / std[i];
                 }
             }
         }
@@ -61,8 +61,8 @@ SimpleTensor<T> normalize_planar_yuv_layer(const SimpleTensor<T> &src, const Sim
     return result;
 }
 
-template SimpleTensor<half> normalize_planar_yuv_layer(const SimpleTensor<half> &src, const SimpleTensor<half> &mean, const SimpleTensor<half> &sd);
-
+template SimpleTensor<half> normalize_planar_yuv_layer(const SimpleTensor<half> &src, const SimpleTensor<half> &mean, const SimpleTensor<half> &std);
+template SimpleTensor<float> normalize_planar_yuv_layer(const SimpleTensor<float> &src, const SimpleTensor<float> &mean, const SimpleTensor<float> &std);
 } // namespace reference
 } // namespace validation
 } // namespace test
diff --git a/tests/validation/reference/NormalizePlanarYUVLayer.h b/tests/validation/reference/NormalizePlanarYUVLayer.h
index c8740a383b..41ce48630c 100644
--- a/tests/validation/reference/NormalizePlanarYUVLayer.h
+++ b/tests/validation/reference/NormalizePlanarYUVLayer.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -36,7 +36,7 @@ namespace validation
 namespace reference
 {
 template <typename T, typename std::enable_if<is_floating_point<T>::value, int>::type * = nullptr>
-SimpleTensor<T> normalize_planar_yuv_layer(const SimpleTensor<T> &src, const SimpleTensor<T> &mean, const SimpleTensor<T> &sd);
+SimpleTensor<T> normalize_planar_yuv_layer(const SimpleTensor<T> &src, const SimpleTensor<T> &mean, const SimpleTensor<T> &std);
 } // namespace reference
 } // namespace validation
 } // namespace test