COMPMID-363 Add Graph library support

Change-Id: Ie841419bf65d0e06bdfe0bdd2d8d4e0bb3631e54
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/87931
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>

164 files changed, 2811 insertions, 314 deletions

diff --git a/SConscript b/SConscript
index d9ee169b2d..00c366e890 100644
--- a/SConscript
+++ b/SConscript
@@ -192,6 +192,25 @@ if env['os'] != 'bare_metal' and not env['standalone']:
     arm_compute_so = build_library('arm_compute', shared_core_objects + shared_runtime_objects, static=False)
     Export('arm_compute_so')
 
+if env['neon'] and env['opencl']:
+    graph_files = Glob('src/graph/*.cpp')
+    graph_files += Glob('src/graph/nodes/*.cpp')
+
+    graph_files += Glob('src/graph/CL/*.cpp')
+    graph_files += Glob('src/graph/NEON/*.cpp')
+
+    shared_graph_objects = [arm_compute_env.SharedObject(f) for f in graph_files]
+    static_graph_objects = [arm_compute_env.StaticObject(f) for f in graph_files]
+
+    arm_compute_graph_a = build_library('arm_compute_graph-static', static_core_objects + static_runtime_objects + static_graph_objects, static=True)
+    Export('arm_compute_graph_a')
+
+    arm_compute_graph_so = build_library('arm_compute_graph', shared_core_objects + shared_runtime_objects + shared_graph_objects, static=False)
+    Export('arm_compute_graph_so')
+
+    graph_alias = arm_compute_env.Alias("arm_compute_graph", [arm_compute_graph_a, arm_compute_graph_so])
+    Default(graph_alias)
+
 if env['standalone']:
     alias = arm_compute_env.Alias("arm_compute", [arm_compute_a])
 else:
diff --git a/arm_compute/graph/CL/CLMap.h b/arm_compute/graph/CL/CLMap.h
new file mode 100644
index 0000000000..a205ebcad1
--- /dev/null
+++ b/arm_compute/graph/CL/CLMap.h
@@ -0,0 +1,66 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_CLMAP_H__
+#define __ARM_COMPUTE_GRAPH_CLMAP_H__
+
+#include "arm_compute/graph/Types.h"
+#include "arm_compute/runtime/IFunction.h"
+
+namespace arm_compute
+{
+class CLTensor;
+
+namespace graph
+{
+class Tensor;
+/** OpenCL map function */
+class CLMap : public arm_compute::IFunction
+{
+public:
+    /** Constructor
+     *
+     * @param[in] tensor   Tensor to map
+     * @param[in] blocking Flag to specify if the map should be blocking or not (defaults to false)
+     */
+    CLMap(Tensor *tensor, bool blocking = false);
+    /** Prevent instances from being copy constructed */
+    CLMap(const CLMap &) = delete;
+    /** Prevent instances from being copy assigned */
+    const CLMap &operator=(const CLMap &) = delete;
+    /** Allow instances to be move constructed */
+    CLMap(CLMap &&) = default;
+    /** Allow instances to be move assigned */
+    CLMap &operator=(CLMap &&) = default;
+
+    // Inherited methods overriden:
+    void run() override;
+
+private:
+    arm_compute::CLTensor *_tensor;   /**< Tensor */
+    bool                   _blocking; /**< Blocking flag */
+};
+} // namespace graph
+} // namespace arm_compute
+
+#endif /* __ARM_COMPUTE_GRAPH_CLMAP_H__ */
diff --git a/arm_compute/graph/CL/CLUnmap.h b/arm_compute/graph/CL/CLUnmap.h
new file mode 100644
index 0000000000..a72706353b
--- /dev/null
+++ b/arm_compute/graph/CL/CLUnmap.h
@@ -0,0 +1,63 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_CLUNMAP_H__
+#define __ARM_COMPUTE_GRAPH_CLUNMAP_H__
+
+#include "arm_compute/graph/Types.h"
+#include "arm_compute/runtime/IFunction.h"
+
+namespace arm_compute
+{
+class CLTensor;
+
+namespace graph
+{
+class Tensor;
+/** OpenCL un-map function */
+class CLUnmap : public arm_compute::IFunction
+{
+public:
+    /** Constructor
+     *
+     * @param[in] tensor Tensor to un-map
+     */
+    CLUnmap(Tensor *tensor);
+    /** Prevent instances from being copy constructed */
+    CLUnmap(const CLUnmap &) = delete;
+    /** Prevent instances from being copy assigned */
+    const CLUnmap &operator=(const CLUnmap &) = delete;
+    /** Allow instances to be move constructed */
+    CLUnmap(CLUnmap &&) = default;
+    /** Allow instances to be move assigned */
+    CLUnmap &operator=(CLUnmap &&) = default;
+
+    // Inherited methods overriden:
+    void run() override;
+
+private:
+    arm_compute::CLTensor *_tensor; /**< Tensor */
+};
+} // namespace graph
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_GRAPH_CLUNMAP_H__ */
diff --git a/arm_compute/graph/Graph.h b/arm_compute/graph/Graph.h
new file mode 100644
index 0000000000..3c263c2bdd
--- /dev/null
+++ b/arm_compute/graph/Graph.h
@@ -0,0 +1,132 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_GRAPH_H__
+#define __ARM_COMPUTE_GRAPH_GRAPH_H__
+
+#include "arm_compute/graph/INode.h"
+#include "arm_compute/graph/Tensor.h"
+#include "arm_compute/graph/Types.h"
+#include "support/ToolchainSupport.h"
+
+#include <memory>
+
+namespace arm_compute
+{
+class IFunction;
+
+namespace graph
+{
+/** Graph class */
+class Graph final
+{
+public:
+    /** Constructor */
+    Graph();
+    /** Destructor */
+    ~Graph();
+    /** Prevent instances from being copy constructed */
+    Graph(const Graph &) = delete;
+    /** Prevent instances from being copy assigned */
+    const Graph &operator=(const Graph &) = delete;
+    /** Prevent instances from being move constructed */
+    Graph(Graph &&) = delete;
+    /** Prevent instances from being move assigned */
+    Graph &operator=(Graph &&) = delete;
+    /** Executes the graph */
+    void run();
+    /** Adds a node to the graph
+     *
+     * @param[in] node Node to add
+     */
+    void add_node(std::unique_ptr<INode> node);
+    /** Adds a tensor to the graph
+     *
+     * @param[in] tensor Tensor to add
+     */
+    void add_tensor(std::unique_ptr<Tensor> tensor);
+    /** Sets an execution hint to the graph
+     *
+     * @note Hint is propagated to the following node and as per name
+     *       its just a hint/preference to be considered by the graph executor
+     *
+     * @param[in] hint execution hint
+     */
+    void set_hint(Hint hint);
+    /** Manually sets the output of the current node
+     *
+     * @param[in] tmp Output info to set
+     */
+    void set_temp(TensorInfo &&tmp);
+
+    /** Sets whether to enable information print out
+     *
+     * @param[in] is_enabled Set to true if need info printed out
+     */
+    void set_info_enablement(bool is_enabled);
+
+private:
+    class Private;
+    std::unique_ptr<Private> _pimpl; /**< Internal implementation class */
+};
+
+/** Overloaded stream operator to add a tensor through its tensor info to the graph
+ *
+ * @param[in, out] graph Graph to add the tensor
+ * @param[in]      info  Tensor information of the tensor to be added
+ *
+ * @return Updated graph
+ */
+Graph &operator<<(Graph &graph, TensorInfo &&info);
+/** Overloaded stream operator to add a tensor to the graph
+ *
+ * @param[in, out] graph  Graph to add the tensor
+ * @param[in]      tensor Tensor to be added
+ *
+ * @return Updated graph
+ */
+Graph &operator<<(Graph &graph, Tensor &&tensor);
+/** Overloaded stream operator to provide an execution hint to the graph
+ *
+ * @param[in, out] graph Graph to provide the hint to
+ * @param[in]      hint  Execution hint to be considered
+ *
+ * @return Updated graph
+ */
+Graph &operator<<(Graph &graph, Hint hint);
+/** Overloaded stream operator to add a node to the graph
+ *
+ * @param[in, out] graph Graph to add the tensor
+ * @param[in]      node  Node to be added
+ *
+ * @return Updated graph
+ */
+template <typename Node>
+Graph &operator<<(Graph &graph, Node node)
+{
+    graph.add_node(arm_compute::support::cpp14::make_unique<Node>(std::move(node)));
+    return graph;
+}
+} // namespace graph
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_GRAPH_GRAPH_H__ */
diff --git a/arm_compute/graph/INode.h b/arm_compute/graph/INode.h
new file mode 100644
index 0000000000..13b5d05f87
--- /dev/null
+++ b/arm_compute/graph/INode.h
@@ -0,0 +1,78 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_INODE_H__
+#define __ARM_COMPUTE_GRAPH_INODE_H__
+
+#include "arm_compute/graph/Types.h"
+#include "arm_compute/runtime/IFunction.h"
+
+#include <memory>
+
+namespace arm_compute
+{
+namespace graph
+{
+/** Node interface */
+class INode
+{
+public:
+    /** Virtual Destructor */
+    virtual ~INode() = default;
+    /** Interface to be implemented that instantiates the node
+     *
+     * @param[in] hint   Hint to where the node should be executed
+     * @param[in] input  Input tensor of the node
+     * @param[in] output Output tensor of the node
+     */
+    virtual std::unique_ptr<arm_compute::IFunction> instantiate_node(Hint hint, ITensor *input, ITensor *output) = 0;
+    /** Override the existing hint
+     *
+     * @note If the input is DONT_CARE then the method has to pick a technology,
+     *       else it can accept the hint or override it (But not with DONT_CARE)
+     *
+     * @param[in] hint Hint to be considered
+     *
+     * @return The updated hint
+     */
+    Hint override_hint(Hint hint) const;
+
+    virtual void print_info() = 0;
+
+protected:
+    /** Interface to be implement that override the hint
+     *
+     * @param[in] hint Hint to be considered
+     *
+     * @return The updated hint
+     */
+    virtual Hint node_override_hint(Hint hint) const;
+
+protected:
+    Hint     _hint{ Hint::DONT_CARE };
+    ITensor *_input{ nullptr };
+    ITensor *_output{ nullptr };
+};
+} // namespace graph
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_GRAPH_INODE_H__ */
diff --git a/arm_compute/graph/ITensorAccessor.h b/arm_compute/graph/ITensorAccessor.h
new file mode 100644
index 0000000000..d6a254ab73
--- /dev/null
+++ b/arm_compute/graph/ITensorAccessor.h
@@ -0,0 +1,49 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_ITENSORACCESSOR_H__
+#define __ARM_COMPUTE_GRAPH_ITENSORACCESSOR_H__
+
+#include "arm_compute/graph/Types.h"
+
+namespace arm_compute
+{
+namespace graph
+{
+/** Tensor accessor interface */
+class ITensorAccessor
+{
+public:
+    /** Default virtual destructor */
+    virtual ~ITensorAccessor() = default;
+    /** Interface to be implemented to access a given tensor
+     *
+     * @param[in] tensor Tensor to be accessed
+     *
+     * @return True if access is successful else false
+     */
+    virtual bool access_tensor(ITensor &tensor) = 0;
+};
+} // namespace graph
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_GRAPH_ITENSORACCESSOR_H__ */
diff --git a/arm_compute/graph/Nodes.h b/arm_compute/graph/Nodes.h
new file mode 100644
index 0000000000..5e995ac8d1
--- /dev/null
+++ b/arm_compute/graph/Nodes.h
@@ -0,0 +1,33 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_NODES_H__
+#define __ARM_COMPUTE_GRAPH_NODES_H__
+
+#include "arm_compute/graph/nodes/ActivationLayer.h"
+#include "arm_compute/graph/nodes/ConvolutionLayer.h"
+#include "arm_compute/graph/nodes/FullyConnectedLayer.h"
+#include "arm_compute/graph/nodes/PoolingLayer.h"
+#include "arm_compute/graph/nodes/SoftmaxLayer.h"
+
+#endif /* __ARM_COMPUTE_GRAPH_NODES_H__ */
diff --git a/arm_compute/graph/Tensor.h b/arm_compute/graph/Tensor.h
new file mode 100644
index 0000000000..0e823ffad0
--- /dev/null
+++ b/arm_compute/graph/Tensor.h
@@ -0,0 +1,127 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_TENSOR_H__
+#define __ARM_COMPUTE_GRAPH_TENSOR_H__
+
+#include "arm_compute/graph/ITensorAccessor.h"
+#include "arm_compute/graph/Types.h"
+#include "support/ToolchainSupport.h"
+
+#include <memory>
+
+namespace arm_compute
+{
+namespace graph
+{
+/** Tensor class */
+class Tensor
+{
+public:
+    /** Constructor
+     *
+     * @param[in] info Tensor info to use
+     */
+    Tensor(TensorInfo &&info);
+    /** Constructor
+     *
+     * @param[in] accessor Tensor accessor
+     */
+    template <typename AccessorType>
+    Tensor(std::unique_ptr<AccessorType> accessor)
+        : _target(Hint::DONT_CARE), _info(), _accessor(std::move(accessor)), _tensor(nullptr)
+    {
+    }
+    /** Constructor
+     *
+     * @param[in] accessor Tensor accessor
+     */
+    template <typename AccessorType>
+    Tensor(AccessorType &&accessor)
+        : _target(Hint::DONT_CARE), _info(), _accessor(arm_compute::support::cpp14::make_unique<AccessorType>(std::forward<AccessorType>(accessor))), _tensor(nullptr)
+    {
+    }
+    /** Constructor
+     *
+     * @param[in] info     Tensor info to use
+     * @param[in] accessor Tensor accessor
+     */
+    template <typename AccessorType>
+    Tensor(TensorInfo &&info, AccessorType &&accessor)
+        : _target(Hint::DONT_CARE), _info(info), _accessor(arm_compute::support::cpp14::make_unique<AccessorType>(std::forward<AccessorType>(accessor))), _tensor(nullptr)
+    {
+    }
+    /** Default Destructor */
+    ~Tensor() = default;
+    /** Move Constructor
+     *
+     * @param[in] src Tensor to move
+     */
+    Tensor(Tensor &&src) noexcept;
+
+    /** Sets the given TensorInfo to the tensor
+     *
+     * @param[in] info TensorInfo to set
+     */
+    void set_info(TensorInfo &&info);
+    /** Calls accessor on tensor
+     *
+     * @return True if succeeds else false
+     */
+    bool call_accessor();
+    /** Sets target of the tensor
+     *
+     * @param[in] target Target where the tensor should be pinned in
+     *
+     * @return
+     */
+    ITensor *set_target(Hint target);
+    /** Returns tensor's TensorInfo
+     *
+     * @return TensorInfo of the tensor
+     */
+    const TensorInfo &info() const;
+    /** Returns a pointer to the internal tensor
+     *
+     * @return Tensor
+     */
+    ITensor *tensor();
+    /** Allocates and fills the tensor if needed */
+    void allocate_and_fill_if_needed();
+    /** Allocates the tensor */
+    void allocate();
+    /** Return the target that this tensor is pinned on
+     *
+     * @return Target of the tensor
+     */
+    Hint target() const;
+
+private:
+    Hint                             _target;   /**< Target that this tensor is pinned on */
+    TensorInfo                       _info;     /**< Tensor metadata */
+    std::unique_ptr<ITensorAccessor> _accessor; /**< Tensor Accessor */
+    std::unique_ptr<ITensor>         _tensor;   /**< Tensor */
+};
+} // namespace graph
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_GRAPH_TENSOR_H__ */
diff --git a/arm_compute/graph/Types.h b/arm_compute/graph/Types.h
new file mode 100644
index 0000000000..0b9596d589
--- /dev/null
+++ b/arm_compute/graph/Types.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_TYPES_H__
+#define __ARM_COMPUTE_GRAPH_TYPES_H__
+
+#include "arm_compute/core/ITensor.h"
+#include "arm_compute/core/TensorInfo.h"
+
+namespace arm_compute
+{
+namespace graph
+{
+using arm_compute::ActivationLayerInfo;
+using arm_compute::ITensor;
+using arm_compute::TensorInfo;
+using arm_compute::DataType;
+using arm_compute::TensorShape;
+using arm_compute::PadStrideInfo;
+using arm_compute::WeightsInfo;
+using arm_compute::PoolingLayerInfo;
+using arm_compute::PoolingType;
+
+/**< Execution hint to the graph executor */
+enum class Hint
+{
+    DONT_CARE, /**< Run node in any device */
+    OPENCL,    /**< Run node on an OpenCL capable device (GPU) */
+    NEON       /**< Run node on a NEON capable device */
+};
+
+} // namespace graph
+} // namespace arm_compute
+#endif /*__ARM_COMPUTE_GRAPH_TYPES_H__*/
diff --git a/arm_compute/graph/nodes/ActivationLayer.h b/arm_compute/graph/nodes/ActivationLayer.h
new file mode 100644
index 0000000000..c23674e7b6
--- /dev/null
+++ b/arm_compute/graph/nodes/ActivationLayer.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_ACTIVATION_LAYER_H__
+#define __ARM_COMPUTE_GRAPH_ACTIVATION_LAYER_H__
+
+#include "arm_compute/graph/INode.h"
+#include "arm_compute/graph/Tensor.h"
+#include "arm_compute/graph/Types.h"
+
+namespace arm_compute
+{
+namespace graph
+{
+/** Activation Layer node */
+class ActivationLayer : public INode
+{
+public:
+    /** Default Constructor
+     *
+     * @param[in] activation_info Activation layer info
+     */
+    ActivationLayer(const ActivationLayerInfo activation_info);
+
+    // Inherited methods overriden:
+    std::unique_ptr<arm_compute::IFunction> instantiate_node(Hint hint, ITensor *input, ITensor *output) override;
+    void print_info() override;
+
+private:
+    const ActivationLayerInfo _activation_info; /**< Activation layer info */
+};
+} // namespace graph
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_GRAPH_ACTIVATION_LAYER_H__ */
diff --git a/arm_compute/graph/nodes/ConvolutionLayer.h b/arm_compute/graph/nodes/ConvolutionLayer.h
new file mode 100644
index 0000000000..c0e257bf6a
--- /dev/null
+++ b/arm_compute/graph/nodes/ConvolutionLayer.h
@@ -0,0 +1,71 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_CONVOLUTION_LAYER_H__
+#define __ARM_COMPUTE_GRAPH_CONVOLUTION_LAYER_H__
+
+#include "arm_compute/graph/INode.h"
+#include "arm_compute/graph/Tensor.h"
+#include "arm_compute/graph/Types.h"
+
+namespace arm_compute
+{
+namespace graph
+{
+/** Convolution layer node */
+class ConvolutionLayer : public INode
+{
+public:
+    /** Default Constructor
+     *
+     * @param[in] conv_width   Convolution width
+     * @param[in] conv_height  Convolution height
+     * @param[in] ofm          Output feature map
+     * @param[in] weights      Weights of the convolution layer
+     * @param[in] biases       Bias of the convolution layer
+     * @param[in] conv_info    Convolution information
+     * @param[in] weights_info Weights information
+     */
+    template <typename AccessorTypeWeights, typename AccessorTypeBiases>
+    ConvolutionLayer(unsigned int conv_width, unsigned int conv_height, unsigned int ofm, AccessorTypeWeights &&weights,
+                     AccessorTypeBiases &&biases, const PadStrideInfo &conv_info, const WeightsInfo &weights_info = WeightsInfo())
+        : _conv_width(conv_width), _conv_height(conv_height), _ofm(ofm), _weights(std::move(weights)), _biases(std::move(biases)), _conv_info(conv_info), _weights_info(weights_info)
+    {
+    }
+
+    // Inherited methods overriden:
+    std::unique_ptr<arm_compute::IFunction> instantiate_node(Hint hint, ITensor *input, ITensor *output) override;
+    void print_info() override;
+
+private:
+    unsigned int         _conv_width;   /**< Convolution width */
+    unsigned int         _conv_height;  /**< Convolution height */
+    unsigned int         _ofm;          /**< Output feature maps */
+    Tensor               _weights;      /**< Weights tensor */
+    Tensor               _biases;       /**< Biases tensor */
+    const PadStrideInfo &_conv_info;    /**< Convolution layer information */
+    const WeightsInfo   &_weights_info; /**< Convolution layer weights information */
+};
+} // namespace graph
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_GRAPH_CONVOLUTION_LAYER_H__ */
diff --git a/arm_compute/graph/nodes/FullyConnectedLayer.h b/arm_compute/graph/nodes/FullyConnectedLayer.h
new file mode 100644
index 0000000000..3e1fe23b11
--- /dev/null
+++ b/arm_compute/graph/nodes/FullyConnectedLayer.h
@@ -0,0 +1,63 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_FULLY_CONNECTED_LAYER_H__
+#define __ARM_COMPUTE_GRAPH_FULLY_CONNECTED_LAYER_H__
+
+#include "arm_compute/graph/INode.h"
+#include "arm_compute/graph/Tensor.h"
+#include "arm_compute/graph/Types.h"
+
+namespace arm_compute
+{
+namespace graph
+{
+/** Fully connected layer node */
+class FullyConnectedLayer : public INode
+{
+public:
+    /** Default constructor
+     *
+     * @param[in] num_neurons Number of neurons
+     * @param[in] weights     Weights of the fully connected layer
+     * @param[in] biases      Biases of the fully connected layer
+     */
+    template <typename AccessorTypeWeights, typename AccessorTypeBiases>
+    FullyConnectedLayer(unsigned int num_neurons, AccessorTypeWeights &&weights, AccessorTypeBiases &&biases)
+        : _num_neurons(num_neurons), _weights(std::move(weights)), _biases(std::move(biases))
+    {
+    }
+
+    // Inherited methods overriden:
+    std::unique_ptr<arm_compute::IFunction> instantiate_node(Hint hint, ITensor *input, ITensor *output) override;
+    void print_info() override;
+
+    // Inherited methods overriden:
+private:
+    unsigned int _num_neurons; /**< Number of neurons */
+    Tensor       _weights;     /**< Weights tensor */
+    Tensor       _biases;      /**< Biases tensor */
+};
+} // namespace graph
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_GRAPH_FULLY_CONNECTED_LAYER_H__ */
diff --git a/arm_compute/graph/nodes/PoolingLayer.h b/arm_compute/graph/nodes/PoolingLayer.h
new file mode 100644
index 0000000000..14e2c6d264
--- /dev/null
+++ b/arm_compute/graph/nodes/PoolingLayer.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_POOLING_LAYER_H__
+#define __ARM_COMPUTE_GRAPH_POOLING_LAYER_H__
+
+#include "arm_compute/graph/INode.h"
+#include "arm_compute/graph/Tensor.h"
+#include "arm_compute/graph/Types.h"
+
+namespace arm_compute
+{
+namespace graph
+{
+/** Pooling layer node */
+class PoolingLayer : public INode
+{
+public:
+    /** Default Constructor
+     *
+     * @param pool_info Pooling layer information
+     */
+    PoolingLayer(const PoolingLayerInfo pool_info);
+
+    // Inherited methods overriden:
+    std::unique_ptr<arm_compute::IFunction> instantiate_node(Hint hint, ITensor *input, ITensor *output) override;
+    void print_info() override;
+
+private:
+    const PoolingLayerInfo _pool_info; /**< Pooling layer information */
+};
+} // namespace graph
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_GRAPH_POOLING_LAYER_H__ */
diff --git a/arm_compute/graph/nodes/SoftmaxLayer.h b/arm_compute/graph/nodes/SoftmaxLayer.h
new file mode 100644
index 0000000000..1779adae66
--- /dev/null
+++ b/arm_compute/graph/nodes/SoftmaxLayer.h
@@ -0,0 +1,46 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_SOFTMAX_LAYER_H__
+#define __ARM_COMPUTE_GRAPH_SOFTMAX_LAYER_H__
+
+#include "arm_compute/graph/INode.h"
+#include "arm_compute/graph/Tensor.h"
+#include "arm_compute/graph/Types.h"
+
+namespace arm_compute
+{
+namespace graph
+{
+/** Softmax layer node */
+class SoftmaxLayer : public INode
+{
+public:
+    // Inherited methods overriden:
+    std::unique_ptr<arm_compute::IFunction> instantiate_node(Hint hint, ITensor *input, ITensor *output) override;
+    void print_info() override;
+};
+
+} // namespace graph
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_GRAPH_SOFTMAX_LAYER_H__ */
diff --git a/examples/SConscript b/examples/SConscript
index 2303109bf5..853a1bb514 100644
--- a/examples/SConscript
+++ b/examples/SConscript
@@ -52,6 +52,20 @@ if env['opencl'] and env['neon']:
         Depends(prog, [arm_compute_dependency, opencl])
         alias = examples_env.Alias(example, prog)
         Default(alias)
+    Import('arm_compute_graph_a')
+    Import('arm_compute_graph_so')
+    if env['os'] == 'android':
+        arm_compute_graph_lib = arm_compute_graph_a
+    else:
+        arm_compute_graph_lib = "arm_compute_graph"
+
+    graph_utils = examples_env.Object("../utils/GraphUtils.cpp")
+    for file in Glob("./graph_*.cpp"):
+        example = os.path.basename(os.path.splitext(str(file))[0])
+        prog = examples_env.Program(example, ["{}.cpp".format(example), utils, graph_utils], CPPDEFINES=['ARM_COMPUTE_CL'], LIBS = [arm_compute_graph_lib, "OpenCL"])
+        Depends(prog, [arm_compute_dependency, opencl])
+        alias = examples_env.Alias(example, prog)
+        Default(alias)
 
 if env['opencl']:
     for file in Glob("./cl_*.cpp"):
diff --git a/examples/graph_lenet.cpp b/examples/graph_lenet.cpp
new file mode 100644
index 0000000000..9e6551973b
--- /dev/null
+++ b/examples/graph_lenet.cpp
@@ -0,0 +1,143 @@
+/*
+ * Copyright (c) 2017 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_CL /* Needed by Utils.cpp to handle OpenCL exceptions properly */
+#error "This example needs to be built with -DARM_COMPUTE_CL"
+#endif /* ARM_COMPUTE_CL */
+
+#include "arm_compute/graph/Graph.h"
+#include "arm_compute/graph/Nodes.h"
+#include "arm_compute/runtime/CL/CLScheduler.h"
+#include "arm_compute/runtime/Scheduler.h"
+#include "support/ToolchainSupport.h"
+#include "utils/GraphUtils.h"
+#include "utils/Utils.h"
+
+#include <cstdlib>
+#include <iostream>
+#include <memory>
+
+using namespace arm_compute::graph;
+using namespace arm_compute::graph_utils;
+
+/** Generates appropriate accessor according to the specified path
+ *
+ * @note If path is empty will generate a DummyAccessor else will generate a NumPyBinLoader
+ *
+ * @param path       Path to the data files
+ * @param data_file  Relative path to the data files from path
+ *
+ * @return An appropriate tensor accessor
+ */
+std::unique_ptr<ITensorAccessor> get_accessor(const std::string &path, const std::string &data_file)
+{
+    if(path.empty())
+    {
+        return arm_compute::support::cpp14::make_unique<DummyAccessor>();
+    }
+    else
+    {
+        return arm_compute::support::cpp14::make_unique<NumPyBinLoader>(path + data_file);
+    }
+}
+
+/** Example demonstrating how to implement LeNet's network using the Compute Library's graph API
+ *
+ * @param[in] argc Number of arguments
+ * @param[in] argv Arguments ( [optional] Path to the weights folder, [optional] batches )
+ */
+void main_graph_lenet(int argc, const char **argv)
+{
+    std::string  data_path;   /** Path to the trainable data */
+    unsigned int batches = 4; /** Number of batches */
+
+    // Parse arguments
+    if(argc < 2)
+    {
+        // Print help
+        std::cout << "Usage: " << argv[0] << " [path_to_data] [batches]\n\n";
+        std::cout << "No data folder provided: using random values\n\n";
+    }
+    else if(argc == 2)
+    {
+        //Do something with argv[1]
+        data_path = argv[1];
+        std::cout << "Usage: " << argv[0] << " [path_to_data] [batches]\n\n";
+        std::cout << "No number of batches where specified, thus will use the default : " << batches << "\n\n";
+    }
+    else
+    {
+        //Do something with argv[1] and argv[2]
+        data_path = argv[1];
+        batches   = std::strtol(argv[2], nullptr, 0);
+    }
+
+    // Check if OpenCL is available and initialize the scheduler
+    if(arm_compute::opencl_is_available())
+    {
+        arm_compute::CLScheduler::get().default_init();
+    }
+
+    Graph graph;
+    graph.set_info_enablement(true);
+
+    //conv1 << pool1 << conv2 << pool2 << fc1 << act1 << fc2 << smx
+    graph << Hint::OPENCL
+          << Tensor(TensorInfo(TensorShape(28U, 28U, 1U, batches), 1, DataType::F32), DummyAccessor())
+          << ConvolutionLayer(
+              5U, 5U, 20U,
+              get_accessor(data_path, "/cnn_data/lenet_model/conv1_w.npy"),
+              get_accessor(data_path, "/cnn_data/lenet_model/conv1_b.npy"),
+              PadStrideInfo(1, 1, 0, 0))
+          << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 2, PadStrideInfo(2, 2, 0, 0)))
+          << ConvolutionLayer(
+              5U, 5U, 50U,
+              get_accessor(data_path, "/cnn_data/lenet_model/conv2_w.npy"),
+              get_accessor(data_path, "/cnn_data/lenet_model/conv2_b.npy"),
+              PadStrideInfo(1, 1, 0, 0))
+          << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 2, PadStrideInfo(2, 2, 0, 0)))
+          << FullyConnectedLayer(
+              500U,
+              get_accessor(data_path, "/cnn_data/lenet_model/ip1_w.npy"),
+              get_accessor(data_path, "/cnn_data/lenet_model/ip1_b.npy"))
+          << TensorInfo(TensorShape(500U, batches), 1, DataType::F32)
+          << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
+          << FullyConnectedLayer(
+              10U,
+              get_accessor(data_path, "/cnn_data/lenet_model/ip2_w.npy"),
+              get_accessor(data_path, "/cnn_data/lenet_model/ip2_b.npy"))
+          << SoftmaxLayer()
+          << Tensor(DummyAccessor());
+
+    graph.run();
+}
+
+/** Main program for LeNet
+ *
+ * @param[in] argc Number of arguments
+ * @param[in] argv Arguments ( [optional] Path to the weights folder, [optional] batches )
+ */
+int main(int argc, const char **argv)
+{
+    return arm_compute::utils::run_example(argc, argv, main_graph_lenet);
+}
diff --git a/scripts/clang_tidy_rules.py b/scripts/clang_tidy_rules.py
index e4daceb071..c3faf8736c 100755
--- a/scripts/clang_tidy_rules.py
+++ b/scripts/clang_tidy_rules.py
@@ -49,6 +49,8 @@ def filter_clang_tidy_lines( lines ):
                 ("omp.h" in line) or
                 ("memory" in line and "cast from pointer to smaller type 'uintptr_t' (aka 'unsigned int') loses information" in line) or
                 ("NEMath.inl" in line and "statement expression not allowed at file scope" in line) or
+                ("Utils.h" in line and "no member named 'unmap' in 'arm_compute::Tensor'" in line) or
+                ("Utils.h" in line and "no member named 'map' in 'arm_compute::Tensor'" in line) or
                 "3rdparty" in line):
                 print_context=False
                 continue
@@ -76,6 +78,7 @@ def filter_clang_tidy_lines( lines ):
                ("NEGEMMMatrixMultiplyKernel.cpp" in line and "do not use C-style cast to convert between unrelated types" in line) or
                ("NEPoolingLayerKernel.cpp" in line and "do not use C-style cast to convert between unrelated types" in line) or
                ("NESoftmaxLayerKernel.cpp" in line and "do not use C-style cast to convert between unrelated types" in line) or
+               ("GraphUtils.cpp" in line and "consider replacing 'unsigned long' with 'uint64'" in line) or
                ("parameter 'memory_manager' is unused" in line) or
                ("parameter 'memory_manager' is copied for each invocation but only used as a const reference" in line) or
                "3rdparty" in line):
diff --git a/src/graph/CL/CLMap.cpp b/src/graph/CL/CLMap.cpp
new file mode 100644
index 0000000000..4892b9609b
--- /dev/null
+++ b/src/graph/CL/CLMap.cpp
@@ -0,0 +1,42 @@
+/*
+ * Copyright (c) 2017 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/graph/CL/CLMap.h"
+
+#include "arm_compute/core/Helpers.h"
+#include "arm_compute/core/Validate.h"
+#include "arm_compute/graph/Tensor.h"
+#include "arm_compute/runtime/CL/CLTensor.h"
+
+using namespace arm_compute::graph;
+
+CLMap::CLMap(Tensor *tensor, bool blocking)
+    : _tensor(dynamic_cast<arm_compute::CLTensor *>(tensor->tensor())), _blocking(blocking)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(_tensor);
+}
+
+void CLMap::run()
+{
+    _tensor->map(_blocking);
+}
diff --git a/src/graph/CL/CLUnmap.cpp b/src/graph/CL/CLUnmap.cpp
new file mode 100644
index 0000000000..ec7d8650d6
--- /dev/null
+++ b/src/graph/CL/CLUnmap.cpp
@@ -0,0 +1,42 @@
+/*
+ * Copyright (c) 2017 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/graph/CL/CLUnmap.h"
+
+#include "arm_compute/core/Helpers.h"
+#include "arm_compute/core/Validate.h"
+#include "arm_compute/graph/Tensor.h"
+#include "arm_compute/runtime/CL/CLTensor.h"
+
+using namespace arm_compute::graph;
+
+CLUnmap::CLUnmap(Tensor *tensor)
+    : _tensor(dynamic_cast<arm_compute::CLTensor *>(tensor->tensor()))
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(_tensor);
+}
+
+void CLUnmap::run()
+{
+    _tensor->unmap();
+}
diff --git a/src/graph/Graph.cpp b/src/graph/Graph.cpp
new file mode 100644
index 0000000000..525506f316
--- /dev/null
+++ b/src/graph/Graph.cpp
@@ -0,0 +1,255 @@
+/*
+ * Copyright (c) 2017 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/graph/Graph.h"
+
+#include "arm_compute/graph/CL/CLMap.h"
+#include "arm_compute/graph/CL/CLUnmap.h"
+#include "arm_compute/graph/INode.h"
+#include "arm_compute/graph/Tensor.h"
+#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/Tensor.h"
+
+using namespace arm_compute::graph;
+
+struct Stage
+{
+    Tensor                                 *_input;
+    Tensor                                 *_output;
+    std::unique_ptr<arm_compute::IFunction> _function;
+};
+
+struct Graph::Private
+{
+public:
+    /** Finalizes the current node's configuration
+     *
+     * @param _next_hint Device execution hint
+     */
+    void configure(Hint _next_hint);
+
+    /** Sets whether to enable information print out
+     *
+     * @param[in] is_enabled Set to true if need info printed out
+     */
+    void set_info_enablement(bool is_enabled);
+
+    std::vector<Stage>                   _pipeline{};
+    std::vector<std::unique_ptr<Tensor>> _tensors{};
+    std::vector<std::unique_ptr<INode>>  _nodes{};
+    Hint                                 _current_hint{ Hint::DONT_CARE };
+    Hint                                 _next_hint{ Hint::DONT_CARE };
+    std::unique_ptr<Tensor>              _graph_input{ nullptr };
+    std::unique_ptr<Tensor>              _graph_output{ nullptr };
+    std::unique_ptr<INode>               _current_node{ nullptr };
+    Tensor                              *_current_output{ nullptr };
+    bool                                 _info_enabled{ false };
+
+private:
+    Tensor *_current_input{ nullptr };
+    Hint    _previous_hint{ Hint::DONT_CARE };
+};
+
+Graph::~Graph() //NOLINT
+{
+    //Can't use =default because the destructor must be defined after Graph::Private's definition
+}
+
+Graph::Graph()
+    : _pimpl{ new Private() }
+{
+}
+
+void Graph::run()
+{
+    while(true)
+    {
+        if(!_pimpl->_graph_input->call_accessor())
+        {
+            return;
+        }
+
+        for(auto &stage : _pimpl->_pipeline)
+        {
+            stage._function->run();
+        }
+
+        if(!_pimpl->_graph_output->call_accessor())
+        {
+            return;
+        }
+    }
+}
+
+//Finalize current node's configuration
+void Graph::Private::configure(Hint _next_hint)
+{
+    ARM_COMPUTE_ERROR_ON(_current_node == nullptr);
+    ARM_COMPUTE_ERROR_ON(_graph_input == nullptr);
+
+    // Is it the first node of the graph ?
+    if(_current_input == nullptr)
+    {
+        _graph_input->set_target(_current_hint);
+        _current_input = _graph_input.get();
+        _previous_hint = _current_hint; // For the first node just assume the previous node was of the same type as this one
+    }
+
+    //Automatic output configuration ?
+    if(_current_output == nullptr)
+    {
+        _tensors.push_back(arm_compute::support::cpp14::make_unique<Tensor>(TensorInfo()));
+        _current_output = _tensors.back().get();
+    }
+
+    // If either the writer or reader node needs OpenCL then use OpenCL memory:
+    if((_next_hint == Hint::OPENCL || _current_hint == Hint::OPENCL))
+    {
+        _current_output->set_target(Hint::OPENCL);
+    }
+    else
+    {
+        _current_output->set_target(Hint::NEON);
+    }
+
+    // Map input if needed
+    std::unique_ptr<arm_compute::IFunction> func = _current_node->instantiate_node(_current_hint, _current_input->tensor(), _current_output->tensor());
+    _current_input->allocate();
+
+    if(_current_input->target() == Hint::OPENCL)
+    {
+        if(_previous_hint == Hint::NEON)
+        {
+            ARM_COMPUTE_ERROR_ON(_current_hint == Hint::NEON);
+            _pipeline.push_back({ _current_input, _current_input, arm_compute::support::cpp14::make_unique<CLUnmap>(_current_input) });
+        }
+        if(_current_hint == Hint::NEON)
+        {
+            ARM_COMPUTE_ERROR_ON(_previous_hint == Hint::NEON);
+            _pipeline.push_back({ _current_input, _current_input, arm_compute::support::cpp14::make_unique<CLMap>(_current_input, true) });
+        }
+    }
+
+    _pipeline.push_back({ _current_input, _current_output, std::move(func) });
+
+    _current_input  = _current_output;
+    _current_output = nullptr;
+    _previous_hint  = _current_hint;
+    _current_hint   = _next_hint;
+}
+
+void Graph::Private::set_info_enablement(bool is_enabled)
+{
+    _info_enabled = is_enabled;
+}
+
+void Graph::add_node(std::unique_ptr<INode> node)
+{
+    ARM_COMPUTE_ERROR_ON_MSG(_pimpl->_graph_input == nullptr, "The graph's input must be set before the first node is added");
+    ARM_COMPUTE_ERROR_ON_MSG(_pimpl->_graph_output != nullptr, "Nothing can be added after the output tensor");
+    //Trigger the creation of the current Node:
+
+    Hint _next_hint = node->override_hint(_pimpl->_next_hint);
+    ARM_COMPUTE_ERROR_ON(_next_hint == Hint::DONT_CARE);
+    if(_pimpl->_current_node)
+    {
+        //Finalize the previous Node:
+        _pimpl->configure(_pimpl->_next_hint);
+
+        if(_pimpl->_info_enabled)
+        {
+            _pimpl->_current_node->print_info();
+        }
+    }
+    else
+    {
+        // If that's the first node then use the same Hint before and after the node.
+        _pimpl->_current_hint = _next_hint;
+    }
+    if(_pimpl->_current_node)
+    {
+        _pimpl->_nodes.push_back(std::move(_pimpl->_current_node));
+    }
+    _pimpl->_current_node = std::move(node);
+}
+void Graph::set_hint(Hint hint)
+{
+    _pimpl->_next_hint = hint;
+}
+
+void Graph::set_info_enablement(bool is_enabled)
+{
+    _pimpl->set_info_enablement(is_enabled);
+}
+
+//Add a tensor with an Accessor (i.e either the input or output of the graph)
+void Graph::add_tensor(std::unique_ptr<Tensor> tensor)
+{
+    // If it's the first Tensor added then it will be the input of the Graph.
+    if(_pimpl->_graph_input == nullptr)
+    {
+        ARM_COMPUTE_ERROR_ON(_pimpl->_graph_output != nullptr);
+        ARM_COMPUTE_ERROR_ON(_pimpl->_current_node != nullptr);
+        _pimpl->_graph_input = std::move(tensor);
+    }
+    else
+    {
+        // Else it will be the output of the Graph
+        ARM_COMPUTE_ERROR_ON(_pimpl->_graph_output != nullptr);
+        ARM_COMPUTE_ERROR_ON(_pimpl->_current_node == nullptr);
+        _pimpl->_graph_output   = std::move(tensor);
+        _pimpl->_current_output = _pimpl->_graph_output.get();
+
+        // Finalize the graph by configuring the last Node of the graph:
+        _pimpl->configure(_pimpl->_current_hint); // Ignore _next_hint as this is the last node, and just use the same hint as before this node.
+        _pimpl->_graph_output->allocate();
+    }
+}
+
+void Graph::set_temp(TensorInfo &&tmp)
+{
+    ARM_COMPUTE_ERROR_ON(_pimpl->_graph_input == nullptr);
+    ARM_COMPUTE_ERROR_ON(_pimpl->_graph_output != nullptr);
+    ARM_COMPUTE_ERROR_ON_MSG(_pimpl->_current_output != nullptr, "TensorInfo for temporary tensor already set");
+
+    _pimpl->_tensors.push_back(arm_compute::support::cpp14::make_unique<Tensor>(std::move(tmp)));
+    _pimpl->_current_output = _pimpl->_tensors.back().get();
+}
+
+Graph &arm_compute::graph::operator<<(Graph &graph, TensorInfo &&info)
+{
+    graph.set_temp(std::move(info));
+    return graph;
+}
+
+Graph &arm_compute::graph::operator<<(Graph &graph, Tensor &&tensor)
+{
+    graph.add_tensor(arm_compute::support::cpp14::make_unique<Tensor>(std::move(tensor)));
+    return graph;
+}
+
+Graph &arm_compute::graph::operator<<(Graph &graph, Hint hint)
+{
+    graph.set_hint(hint);
+    return graph;
+}
diff --git a/src/graph/INode.cpp b/src/graph/INode.cpp
new file mode 100644
index 0000000000..6b25022d15
--- /dev/null
+++ b/src/graph/INode.cpp
@@ -0,0 +1,47 @@
+/*
+ * Copyright (c) 2017 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/graph/INode.h"
+
+#include "arm_compute/core/CL/OpenCL.h"
+#include "arm_compute/core/Validate.h"
+
+#include <ostream>
+
+using namespace arm_compute::graph;
+
+Hint INode::override_hint(Hint hint) const
+{
+    if(hint == Hint::OPENCL && !opencl_is_available())
+    {
+        hint = Hint::DONT_CARE;
+    }
+    hint = node_override_hint(hint);
+    ARM_COMPUTE_ERROR_ON(hint == Hint::OPENCL && !opencl_is_available());
+    return hint;
+}
+Hint INode::node_override_hint(Hint hint) const
+{
+    return hint == Hint::DONT_CARE ? Hint::NEON : hint;
+}
diff --git a/src/graph/Tensor.cpp b/src/graph/Tensor.cpp
new file mode 100644
index 0000000000..c534ae0296
--- /dev/null
+++ b/src/graph/Tensor.cpp
@@ -0,0 +1,151 @@
+/*
+ * Copyright (c) 2017 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/graph/Tensor.h"
+
+#include "arm_compute/core/Error.h"
+#include "arm_compute/core/Helpers.h"
+#include "arm_compute/core/Validate.h"
+#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/Tensor.h"
+#include "utils/TypePrinter.h"
+
+using namespace arm_compute::graph;
+
+namespace
+{
+template <typename TensorType>
+std::unique_ptr<ITensor> initialise_tensor(TensorInfo &info)
+{
+    auto tensor = arm_compute::support::cpp14::make_unique<TensorType>();
+    tensor->allocator()->init(info);
+    return std::move(tensor);
+}
+
+template <typename TensorType>
+void tensor_allocate(ITensor &tensor)
+{
+    auto itensor = dynamic_cast<TensorType *>(&tensor);
+    ARM_COMPUTE_ERROR_ON_NULLPTR(itensor);
+    itensor->allocator()->allocate();
+}
+} // namespace
+
+Tensor::Tensor(TensorInfo &&info)
+    : _target(Hint::DONT_CARE), _info(info), _accessor(nullptr), _tensor(nullptr)
+{
+}
+
+Tensor::Tensor(Tensor &&src) noexcept
+    : _target(src._target),
+      _info(std::move(src._info)),
+      _accessor(std::move(src._accessor)),
+      _tensor(std::move(src._tensor))
+{
+}
+
+void Tensor::set_info(TensorInfo &&info)
+{
+    _info = info;
+}
+
+bool Tensor::call_accessor()
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(_accessor.get());
+    auto cl_tensor = dynamic_cast<arm_compute::CLTensor *>(_tensor.get());
+    if(cl_tensor != nullptr && cl_tensor->buffer() == nullptr)
+    {
+        cl_tensor->map();
+    }
+    bool retval = _accessor->access_tensor(*_tensor);
+    if(cl_tensor != nullptr)
+    {
+        cl_tensor->unmap();
+    }
+    return retval;
+}
+
+ITensor *Tensor::tensor()
+{
+    return _tensor.get();
+}
+
+const TensorInfo &Tensor::info() const
+{
+    return _info;
+}
+
+ITensor *Tensor::set_target(Hint target)
+{
+    if(_tensor != nullptr)
+    {
+        ARM_COMPUTE_ERROR_ON(target != _target);
+    }
+    else
+    {
+        switch(target)
+        {
+            case Hint::OPENCL:
+                _tensor = initialise_tensor<arm_compute::CLTensor>(_info);
+                break;
+            case Hint::NEON:
+                _tensor = initialise_tensor<arm_compute::Tensor>(_info);
+                break;
+            default:
+                ARM_COMPUTE_ERROR("Invalid Hint");
+        }
+        _target = target;
+    }
+    return _tensor.get();
+}
+
+void Tensor::allocate()
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(_tensor.get());
+    switch(_target)
+    {
+        case Hint::OPENCL:
+            tensor_allocate<arm_compute::CLTensor>(*_tensor);
+            break;
+        case Hint::NEON:
+            tensor_allocate<arm_compute::Tensor>(*_tensor);
+            break;
+        default:
+            ARM_COMPUTE_ERROR("Invalid Hint");
+    }
+}
+
+void Tensor::allocate_and_fill_if_needed()
+{
+    allocate();
+    if(_accessor != nullptr)
+    {
+        call_accessor();
+    }
+}
+
+Hint Tensor::target() const
+{
+    return _target;
+}
diff --git a/src/graph/nodes/ActivationLayer.cpp b/src/graph/nodes/ActivationLayer.cpp
new file mode 100644
index 0000000000..b71e22c601
--- /dev/null
+++ b/src/graph/nodes/ActivationLayer.cpp
@@ -0,0 +1,106 @@
+/*
+ * Copyright (c) 2017 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/graph/nodes/ActivationLayer.h"
+
+#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/functions/CLActivationLayer.h"
+#include "arm_compute/runtime/NEON/functions/NEActivationLayer.h"
+#include "arm_compute/runtime/Tensor.h"
+#include "support/ToolchainSupport.h"
+#include "utils/TypePrinter.h"
+
+using namespace arm_compute::graph;
+
+namespace
+{
+template <typename ActivationType, typename TensorType, Hint hint>
+std::unique_ptr<arm_compute::IFunction> instantiate_function(ITensor *input, ITensor *output, const ActivationLayerInfo &activation_info)
+{
+    auto activation = arm_compute::support::cpp14::make_unique<ActivationType>();
+    activation->configure(
+        dynamic_cast<TensorType *>(input),
+        dynamic_cast<TensorType *>(output),
+        activation_info);
+
+    return std::move(activation);
+}
+
+template <Hint                          hint>
+std::unique_ptr<arm_compute::IFunction> instantiate(ITensor *input, ITensor *output, const ActivationLayerInfo &activation_info);
+
+template <>
+std::unique_ptr<arm_compute::IFunction> instantiate<Hint::OPENCL>(ITensor *input, ITensor *output, const ActivationLayerInfo &activation_info)
+{
+    return instantiate_function<arm_compute::CLActivationLayer, arm_compute::CLTensor, Hint::OPENCL>(input, output, activation_info);
+}
+
+template <>
+std::unique_ptr<arm_compute::IFunction> instantiate<Hint::NEON>(ITensor *input, ITensor *output, const ActivationLayerInfo &activation_info)
+{
+    return instantiate_function<arm_compute::NEActivationLayer, arm_compute::Tensor, Hint::NEON>(input, output, activation_info);
+}
+} // namespace
+
+ActivationLayer::ActivationLayer(const ActivationLayerInfo activation_info)
+    : _activation_info(activation_info)
+{
+}
+
+std::unique_ptr<arm_compute::IFunction> ActivationLayer::instantiate_node(Hint hint, ITensor *input, ITensor *output)
+{
+    std::unique_ptr<arm_compute::IFunction> func;
+    _hint   = hint;
+    _input  = input;
+    _output = output;
+
+    if(_hint == Hint::OPENCL)
+    {
+        func = instantiate<Hint::OPENCL>(input, output, _activation_info);
+    }
+    else
+    {
+        func = instantiate<Hint::NEON>(input, output, _activation_info);
+    }
+    return func;
+}
+
+void ActivationLayer::print_info()
+{
+    if(_hint == Hint::OPENCL)
+    {
+        std::cout << "Instantiating CLActivationLayer";
+    }
+    else
+    {
+        std::cout << "Instantiating NEActivationLayer";
+    }
+
+    std::cout << " Data Type: " << _input->info()->data_type()
+              << " Input shape: " << _input->info()->tensor_shape()
+              << " Output shape: " << _output->info()->tensor_shape()
+              << " Activation function: " << _activation_info.activation()
+              << " a: " << _activation_info.a()
+              << " b: " << _activation_info.b()
+              << std::endl;
+}
diff --git a/src/graph/nodes/ConvolutionLayer.cpp b/src/graph/nodes/ConvolutionLayer.cpp
new file mode 100644
index 0000000000..b80bf93eff
--- /dev/null
+++ b/src/graph/nodes/ConvolutionLayer.cpp
@@ -0,0 +1,117 @@
+/*
+ * Copyright (c) 2017 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/graph/nodes/ConvolutionLayer.h"
+
+#include "arm_compute/runtime/CL/functions/CLConvolutionLayer.h"
+#include "arm_compute/runtime/NEON/functions/NEConvolutionLayer.h"
+#include "support/ToolchainSupport.h"
+#include "utils/TypePrinter.h"
+
+using namespace arm_compute::graph;
+
+namespace
+{
+template <typename ConvolutionType, typename TensorType, Hint hint>
+std::unique_ptr<arm_compute::IFunction> instantiate_function(ITensor *input, Tensor &weights, Tensor &biases, ITensor *output, const PadStrideInfo &conv_info, const WeightsInfo &weights_info)
+{
+    bool weights_are_loaded = weights.tensor() != nullptr;
+    bool biases_are_loaded  = biases.tensor() != nullptr;
+
+    auto conv = arm_compute::support::cpp14::make_unique<ConvolutionType>();
+    conv->configure(
+        dynamic_cast<TensorType *>(input),
+        dynamic_cast<TensorType *>(weights.set_target(hint)),
+        dynamic_cast<TensorType *>(biases.set_target(hint)),
+        dynamic_cast<TensorType *>(output),
+        conv_info, weights_info);
+    if(!weights_are_loaded)
+    {
+        weights.allocate_and_fill_if_needed();
+    }
+    if(!biases_are_loaded)
+    {
+        biases.allocate_and_fill_if_needed();
+    }
+
+    return std::move(conv);
+}
+
+template <Hint                          hint>
+std::unique_ptr<arm_compute::IFunction> instantiate(ITensor *input, Tensor &weights, Tensor &biases, ITensor *output, const PadStrideInfo &conv_info, const WeightsInfo &weights_info);
+
+template <>
+std::unique_ptr<arm_compute::IFunction> instantiate<Hint::OPENCL>(ITensor *input, Tensor &weights, Tensor &biases, ITensor *output, const PadStrideInfo &conv_info, const WeightsInfo &weights_info)
+{
+    return instantiate_function<arm_compute::CLConvolutionLayer, arm_compute::CLTensor, Hint::OPENCL>(input, weights, biases, output, conv_info, weights_info);
+}
+
+template <>
+std::unique_ptr<arm_compute::IFunction> instantiate<Hint::NEON>(ITensor *input, Tensor &weights, Tensor &biases, ITensor *output, const PadStrideInfo &conv_info, const WeightsInfo &weights_info)
+{
+    return instantiate_function<arm_compute::NEConvolutionLayer, arm_compute::Tensor, Hint::NEON>(input, weights, biases, output, conv_info, weights_info);
+}
+} // namespace
+
+std::unique_ptr<arm_compute::IFunction> ConvolutionLayer::instantiate_node(Hint hint, ITensor *input, ITensor *output)
+{
+    if(_weights.tensor() == nullptr)
+    {
+        _weights.set_info(TensorInfo(TensorShape(_conv_width, _conv_height, input->info()->dimension(2), _ofm), input->info()->num_channels(), input->info()->data_type(),
+                                     input->info()->fixed_point_position()));
+    }
+    if(_biases.tensor() == nullptr)
+    {
+        _biases.set_info(TensorInfo(TensorShape(_ofm), input->info()->num_channels(), input->info()->data_type(), input->info()->fixed_point_position()));
+    }
+
+    std::unique_ptr<arm_compute::IFunction> func;
+    _hint   = hint;
+    _input  = input;
+    _output = output;
+
+    if(_hint == Hint::OPENCL)
+    {
+        func = instantiate<Hint::OPENCL>(input, _weights, _biases, output, _conv_info, _weights_info);
+    }
+    else
+    {
+        func = instantiate<Hint::NEON>(input, _weights, _biases, output, _conv_info, _weights_info);
+    }
+
+    return func;
+}
+
+void ConvolutionLayer::print_info()
+{
+    if(_hint == Hint::OPENCL)
+    {
+        std::cout << "Instantiating CLConvolutionLayer";
+    }
+    else
+    {
+        std::cout << "Instantiating NEConvolutionLayer";
+    }
+    std::cout << " Type: " << _input->info()->data_type() << " Input Shape: " << _input->info()->tensor_shape() << " Weights shape: " << _weights.info().tensor_shape() << " Biases Shape: " <<
+              _biases.info().tensor_shape() << " Output Shape: " << _output->info()->tensor_shape() << " PadStrideInfo: " << _conv_info << "WeightsInfo: " << _weights_info << std::endl;
+}
diff --git a/src/graph/nodes/FullyConnectedLayer.cpp b/src/graph/nodes/FullyConnectedLayer.cpp
new file mode 100644
index 0000000000..8d244cb515
--- /dev/null
+++ b/src/graph/nodes/FullyConnectedLayer.cpp
@@ -0,0 +1,130 @@
+/*
+ * Copyright (c) 2017 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/graph/nodes/FullyConnectedLayer.h"
+
+#include "arm_compute/core/Helpers.h"
+#include "arm_compute/runtime/CL/functions/CLFullyConnectedLayer.h"
+#include "arm_compute/runtime/NEON/functions/NEFullyConnectedLayer.h"
+#include "support/ToolchainSupport.h"
+#include "utils/TypePrinter.h"
+
+using namespace arm_compute::graph;
+
+namespace
+{
+template <typename FullyConnectedType, typename TensorType, Hint hint>
+std::unique_ptr<arm_compute::IFunction> instantiate_function(ITensor *input, Tensor &weights, Tensor &biases, ITensor *output)
+{
+    bool weights_are_loaded = weights.tensor() != nullptr;
+    bool biases_are_loaded  = biases.tensor() != nullptr;
+
+    auto conv = arm_compute::support::cpp14::make_unique<FullyConnectedType>();
+    conv->configure(
+        dynamic_cast<TensorType *>(input),
+        dynamic_cast<TensorType *>(weights.set_target(hint)),
+        dynamic_cast<TensorType *>(biases.set_target(hint)),
+        dynamic_cast<TensorType *>(output));
+    if(!weights_are_loaded)
+    {
+        weights.allocate_and_fill_if_needed();
+    }
+    if(!biases_are_loaded)
+    {
+        biases.allocate_and_fill_if_needed();
+    }
+
+    return std::move(conv);
+}
+
+template <Hint                          hint>
+std::unique_ptr<arm_compute::IFunction> instantiate(ITensor *input, Tensor &weights, Tensor &biases, ITensor *output);
+
+template <>
+std::unique_ptr<arm_compute::IFunction> instantiate<Hint::OPENCL>(ITensor *input, Tensor &weights, Tensor &biases, ITensor *output)
+{
+    return instantiate_function<arm_compute::CLFullyConnectedLayer, arm_compute::CLTensor, Hint::OPENCL>(input, weights, biases, output);
+}
+
+template <>
+std::unique_ptr<arm_compute::IFunction> instantiate<Hint::NEON>(ITensor *input, Tensor &weights, Tensor &biases, ITensor *output)
+{
+    return instantiate_function<arm_compute::NEFullyConnectedLayer, arm_compute::Tensor, Hint::NEON>(input, weights, biases, output);
+}
+} // namespace
+
+std::unique_ptr<arm_compute::IFunction> FullyConnectedLayer::instantiate_node(Hint hint, ITensor *input, ITensor *output)
+{
+    if(_weights.tensor() == nullptr)
+    {
+        unsigned int num_weights    = 1;
+        unsigned int num_dimensions = input->info()->num_dimensions();
+        // Ignore the batch dimension if there is one:
+        if(num_dimensions == 2 || num_dimensions == 4)
+        {
+            num_dimensions--;
+        }
+        for(unsigned int i = 0; i < num_dimensions; i++)
+        {
+            num_weights *= input->info()->dimension(i);
+        }
+        _weights.set_info(TensorInfo(TensorShape(num_weights, _num_neurons), input->info()->num_channels(), input->info()->data_type(), input->info()->fixed_point_position()));
+    }
+    if(_biases.tensor() == nullptr)
+    {
+        _biases.set_info(TensorInfo(TensorShape(_num_neurons), input->info()->num_channels(), input->info()->data_type(), input->info()->fixed_point_position()));
+    }
+
+    arm_compute::auto_init_if_empty(*output->info(), TensorShape(_num_neurons, input->info()->dimension(1)), input->info()->num_channels(), input->info()->data_type(),
+                                    input->info()->fixed_point_position());
+
+    std::unique_ptr<arm_compute::IFunction> func;
+    _hint   = hint;
+    _input  = input;
+    _output = output;
+
+    if(_hint == Hint::OPENCL)
+    {
+        func = instantiate<Hint::OPENCL>(input, _weights, _biases, output);
+    }
+    else
+    {
+        func = instantiate<Hint::NEON>(input, _weights, _biases, output);
+    }
+
+    return func;
+}
+
+void FullyConnectedLayer::print_info()
+{
+    if(_hint == Hint::OPENCL)
+    {
+        std::cout << "Instantiating CLFullyConnectedLayer";
+    }
+    else
+    {
+        std::cout << "Instantiating NEFullyConnectedLayer";
+    }
+    std::cout << " Type: " << _input->info()->data_type() << " Input Shape: " << _input->info()->tensor_shape() << " Weights shape: " << _weights.info().tensor_shape() << " Biases Shape: " <<
+              _biases.info().tensor_shape() << " Output Shape: " << _output->info()->tensor_shape() << std::endl;
+}
diff --git a/src/graph/nodes/PoolingLayer.cpp b/src/graph/nodes/PoolingLayer.cpp
new file mode 100644
index 0000000000..f29332f65b
--- /dev/null
+++ b/src/graph/nodes/PoolingLayer.cpp
@@ -0,0 +1,104 @@
+/*
+ * Copyright (c) 2017 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/graph/nodes/PoolingLayer.h"
+
+#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/functions/CLPoolingLayer.h"
+#include "arm_compute/runtime/NEON/functions/NEPoolingLayer.h"
+#include "arm_compute/runtime/Tensor.h"
+#include "support/ToolchainSupport.h"
+#include "utils/TypePrinter.h"
+
+using namespace arm_compute::graph;
+
+namespace
+{
+template <typename PoolingType, typename TensorType, Hint hint>
+std::unique_ptr<arm_compute::IFunction> instantiate_function(ITensor *input, ITensor *output, const PoolingLayerInfo &pool_info)
+{
+    auto pool = arm_compute::support::cpp14::make_unique<PoolingType>();
+    pool->configure(
+        dynamic_cast<TensorType *>(input),
+        dynamic_cast<TensorType *>(output),
+        pool_info);
+
+    return std::move(pool);
+}
+
+template <Hint                          hint>
+std::unique_ptr<arm_compute::IFunction> instantiate(ITensor *input, ITensor *output, const PoolingLayerInfo &pool_info);
+
+template <>
+std::unique_ptr<arm_compute::IFunction> instantiate<Hint::OPENCL>(ITensor *input, ITensor *output, const PoolingLayerInfo &pool_info)
+{
+    return instantiate_function<arm_compute::CLPoolingLayer, arm_compute::CLTensor, Hint::OPENCL>(input, output, pool_info);
+}
+
+template <>
+std::unique_ptr<arm_compute::IFunction> instantiate<Hint::NEON>(ITensor *input, ITensor *output, const PoolingLayerInfo &pool_info)
+{
+    return instantiate_function<arm_compute::NEPoolingLayer, arm_compute::Tensor, Hint::NEON>(input, output, pool_info);
+}
+} // namespace
+
+PoolingLayer::PoolingLayer(const PoolingLayerInfo pool_info)
+    : _pool_info(pool_info)
+{
+}
+
+std::unique_ptr<arm_compute::IFunction> PoolingLayer::instantiate_node(Hint hint, ITensor *input, ITensor *output)
+{
+    std::unique_ptr<arm_compute::IFunction> func;
+    _hint   = hint;
+    _input  = input;
+    _output = output;
+
+    if(_hint == Hint::OPENCL)
+    {
+        func = instantiate<Hint::OPENCL>(input, output, _pool_info);
+    }
+    else
+    {
+        func = instantiate<Hint::NEON>(input, output, _pool_info);
+    }
+
+    return func;
+}
+
+void PoolingLayer::print_info()
+{
+    if(_hint == Hint::OPENCL)
+    {
+        std::cout << "Instantiating CLPoolingLayer";
+    }
+    else
+    {
+        std::cout << "Instantiating NEPoolingLayer";
+    }
+
+    std::cout << " Data Type: " << _input->info()->data_type()
+              << " Input shape: " << _input->info()->tensor_shape()
+              << " Output shape: " << _output->info()->tensor_shape()
+              << " Pooling info: " << _pool_info << std::endl;
+}
diff --git a/src/graph/nodes/SoftmaxLayer.cpp b/src/graph/nodes/SoftmaxLayer.cpp
new file mode 100644
index 0000000000..fee88970fc
--- /dev/null
+++ b/src/graph/nodes/SoftmaxLayer.cpp
@@ -0,0 +1,97 @@
+/*
+ * Copyright (c) 2017 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/graph/nodes/SoftmaxLayer.h"
+
+#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/functions/CLSoftmaxLayer.h"
+#include "arm_compute/runtime/NEON/functions/NESoftmaxLayer.h"
+#include "arm_compute/runtime/Tensor.h"
+#include "support/ToolchainSupport.h"
+#include "utils/TypePrinter.h"
+
+using namespace arm_compute::graph;
+
+namespace
+{
+template <typename SoftmaxType, typename TensorType, Hint hint>
+std::unique_ptr<arm_compute::IFunction> instantiate_function(ITensor *input, ITensor *output)
+{
+    auto softmax = arm_compute::support::cpp14::make_unique<SoftmaxType>();
+    softmax->configure(
+        dynamic_cast<TensorType *>(input),
+        dynamic_cast<TensorType *>(output));
+
+    return std::move(softmax);
+}
+
+template <Hint                          hint>
+std::unique_ptr<arm_compute::IFunction> instantiate(ITensor *input, ITensor *output);
+
+template <>
+std::unique_ptr<arm_compute::IFunction> instantiate<Hint::OPENCL>(ITensor *input, ITensor *output)
+{
+    return instantiate_function<arm_compute::CLSoftmaxLayer, arm_compute::CLTensor, Hint::OPENCL>(input, output);
+}
+
+template <>
+std::unique_ptr<arm_compute::IFunction> instantiate<Hint::NEON>(ITensor *input, ITensor *output)
+{
+    return instantiate_function<arm_compute::NESoftmaxLayer, arm_compute::Tensor, Hint::NEON>(input, output);
+}
+} // namespace
+
+std::unique_ptr<arm_compute::IFunction> SoftmaxLayer::instantiate_node(Hint hint, ITensor *input, ITensor *output)
+{
+    std::unique_ptr<arm_compute::IFunction> func;
+    _hint   = hint;
+    _input  = input;
+    _output = output;
+
+    if(_hint == Hint::OPENCL)
+    {
+        func = instantiate<Hint::OPENCL>(input, output);
+    }
+    else
+    {
+        func = instantiate<Hint::NEON>(input, output);
+    }
+
+    return func;
+}
+
+void SoftmaxLayer::print_info()
+{
+    if(_hint == Hint::OPENCL)
+    {
+        std::cout << "Instantiating CLSoftmaxLayer";
+    }
+    else
+    {
+        std::cout << "Instantiating NESoftmaxLayer";
+    }
+    std::cout << " Data Type: " << _input->info()->data_type()
+              << " Input shape: " << _input->info()->tensor_shape()
+              << " Output shape: " << _output->info()->tensor_shape()
+              << std::endl;
+}
diff --git a/tests/AssetsLibrary.cpp b/tests/AssetsLibrary.cpp
index d7c881d9a8..529a2a2d95 100644
--- a/tests/AssetsLibrary.cpp
+++ b/tests/AssetsLibrary.cpp
@@ -23,8 +23,8 @@
  */
 #include "tests/AssetsLibrary.h"
 
-#include "TypePrinter.h"
 #include "Utils.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/ITensor.h"
 
diff --git a/tests/benchmark/CL/ActivationLayer.cpp b/tests/benchmark/CL/ActivationLayer.cpp
index 50a2004b07..e919c7820b 100644
--- a/tests/benchmark/CL/ActivationLayer.cpp
+++ b/tests/benchmark/CL/ActivationLayer.cpp
@@ -27,7 +27,6 @@
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
 #include "arm_compute/runtime/CL/functions/CLActivationLayer.h"
 #include "tests/CL/CLAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/ActivationLayerFixture.h"
 #include "tests/datasets/system_tests/alexnet/AlexNetActivationLayerDataset.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1ActivationLayerDataset.h"
@@ -38,6 +37,7 @@
 #include "tests/datasets/system_tests/yolo/v2/YOLOV2ActivationLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/CL/BatchNormalizationLayer.cpp b/tests/benchmark/CL/BatchNormalizationLayer.cpp
index b0e36d31af..af88278fef 100644
--- a/tests/benchmark/CL/BatchNormalizationLayer.cpp
+++ b/tests/benchmark/CL/BatchNormalizationLayer.cpp
@@ -27,12 +27,12 @@
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
 #include "arm_compute/runtime/CL/functions/CLBatchNormalizationLayer.h"
 #include "tests/CL/CLAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/BatchNormalizationLayerFixture.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4BatchNormalizationLayerDataset.h"
 #include "tests/datasets/system_tests/yolo/v2/YOLOV2BatchNormalizationLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/CL/ConvolutionLayer.cpp b/tests/benchmark/CL/ConvolutionLayer.cpp
index 620feafa83..4467765878 100644
--- a/tests/benchmark/CL/ConvolutionLayer.cpp
+++ b/tests/benchmark/CL/ConvolutionLayer.cpp
@@ -27,7 +27,6 @@
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
 #include "arm_compute/runtime/CL/functions/CLConvolutionLayer.h"
 #include "tests/CL/CLAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/ConvolutionLayerFixture.h"
 #include "tests/datasets/system_tests/alexnet/AlexNetConvolutionLayerDataset.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1ConvolutionLayerDataset.h"
@@ -38,6 +37,7 @@
 #include "tests/datasets/system_tests/yolo/v2/YOLOV2ConvolutionLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/CL/DepthwiseConvolution.cpp b/tests/benchmark/CL/DepthwiseConvolution.cpp
index a8c229fad8..40412da6f9 100644
--- a/tests/benchmark/CL/DepthwiseConvolution.cpp
+++ b/tests/benchmark/CL/DepthwiseConvolution.cpp
@@ -27,11 +27,11 @@
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
 #include "arm_compute/runtime/CL/functions/CLDepthwiseConvolution.h"
 #include "tests/CL/CLAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/DepthwiseConvolutionFixture.h"
 #include "tests/datasets/MobileNetDepthwiseConvolutionDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/CL/DepthwiseSeparableConvolutionLayer.cpp b/tests/benchmark/CL/DepthwiseSeparableConvolutionLayer.cpp
index 0c4fbb18bd..7f5bb00583 100644
--- a/tests/benchmark/CL/DepthwiseSeparableConvolutionLayer.cpp
+++ b/tests/benchmark/CL/DepthwiseSeparableConvolutionLayer.cpp
@@ -27,11 +27,11 @@
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
 #include "arm_compute/runtime/CL/functions/CLDepthwiseSeparableConvolutionLayer.h"
 #include "tests/CL/CLAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/DepthwiseSeparableConvolutionLayerFixture.h"
 #include "tests/datasets/MobileNetDepthwiseSeparableConvolutionLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/CL/DirectConvolutionLayer.cpp b/tests/benchmark/CL/DirectConvolutionLayer.cpp
index 15bf8084ea..2430a4ee94 100644
--- a/tests/benchmark/CL/DirectConvolutionLayer.cpp
+++ b/tests/benchmark/CL/DirectConvolutionLayer.cpp
@@ -27,7 +27,6 @@
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
 #include "arm_compute/runtime/CL/functions/CLDirectConvolutionLayer.h"
 #include "tests/CL/CLAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/ConvolutionLayerFixture.h"
 #include "tests/datasets/system_tests/alexnet/AlexNetConvolutionLayerDataset.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1ConvolutionLayerDataset.h"
@@ -37,6 +36,7 @@
 #include "tests/datasets/system_tests/yolo/v2/YOLOV2ConvolutionLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/CL/Floor.cpp b/tests/benchmark/CL/Floor.cpp
index ce8fa487c1..b2ada0ebc5 100644
--- a/tests/benchmark/CL/Floor.cpp
+++ b/tests/benchmark/CL/Floor.cpp
@@ -27,11 +27,11 @@
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
 #include "arm_compute/runtime/CL/functions/CLFloor.h"
 #include "tests/CL/CLAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/FloorFixture.h"
 #include "tests/datasets/ShapeDatasets.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/CL/FullyConnectedLayer.cpp b/tests/benchmark/CL/FullyConnectedLayer.cpp
index d41424acce..e5f3217368 100644
--- a/tests/benchmark/CL/FullyConnectedLayer.cpp
+++ b/tests/benchmark/CL/FullyConnectedLayer.cpp
@@ -27,7 +27,6 @@
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
 #include "arm_compute/runtime/CL/functions/CLFullyConnectedLayer.h"
 #include "tests/CL/CLAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/FullyConnectedLayerFixture.h"
 #include "tests/datasets/system_tests/alexnet/AlexNetFullyConnectedLayerDataset.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1FullyConnectedLayerDataset.h"
@@ -36,6 +35,7 @@
 #include "tests/datasets/system_tests/vgg/vgg16/VGG16FullyConnectedLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/CL/GEMM.cpp b/tests/benchmark/CL/GEMM.cpp
index 3febef5613..ece20433b9 100644
--- a/tests/benchmark/CL/GEMM.cpp
+++ b/tests/benchmark/CL/GEMM.cpp
@@ -26,12 +26,12 @@
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
 #include "arm_compute/runtime/CL/functions/CLGEMM.h"
 #include "tests/CL/CLAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/GEMMFixture.h"
 #include "tests/datasets/MatrixMultiplyGEMMDataset.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1GEMMDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/CL/NormalizationLayer.cpp b/tests/benchmark/CL/NormalizationLayer.cpp
index 2035ae301a..8dd5f62dd3 100644
--- a/tests/benchmark/CL/NormalizationLayer.cpp
+++ b/tests/benchmark/CL/NormalizationLayer.cpp
@@ -27,12 +27,12 @@
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
 #include "arm_compute/runtime/CL/functions/CLNormalizationLayer.h"
 #include "tests/CL/CLAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/NormalizationLayerFixture.h"
 #include "tests/datasets/system_tests/alexnet/AlexNetNormalizationLayerDataset.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1NormalizationLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/CL/PoolingLayer.cpp b/tests/benchmark/CL/PoolingLayer.cpp
index fcc37e9571..116639a1d0 100644
--- a/tests/benchmark/CL/PoolingLayer.cpp
+++ b/tests/benchmark/CL/PoolingLayer.cpp
@@ -27,7 +27,6 @@
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
 #include "arm_compute/runtime/CL/functions/CLPoolingLayer.h"
 #include "tests/CL/CLAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/PoolingLayerFixture.h"
 #include "tests/datasets/system_tests/alexnet/AlexNetPoolingLayerDataset.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1PoolingLayerDataset.h"
@@ -38,6 +37,7 @@
 #include "tests/datasets/system_tests/yolo/v2/YOLOV2PoolingLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/CL/ROIPoolingLayer.cpp b/tests/benchmark/CL/ROIPoolingLayer.cpp
index a10134fb04..4e78673287 100644
--- a/tests/benchmark/CL/ROIPoolingLayer.cpp
+++ b/tests/benchmark/CL/ROIPoolingLayer.cpp
@@ -29,11 +29,11 @@
 #include "arm_compute/runtime/CL/functions/CLROIPoolingLayer.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/CL/CLArrayAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/ROIPoolingLayerFixture.h"
 #include "tests/datasets/ROIPoolingLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/CL/SYSTEM/AlexNet.cpp b/tests/benchmark/CL/SYSTEM/AlexNet.cpp
index fcf8a52f9e..5f258412de 100644
--- a/tests/benchmark/CL/SYSTEM/AlexNet.cpp
+++ b/tests/benchmark/CL/SYSTEM/AlexNet.cpp
@@ -34,10 +34,10 @@
 #include "arm_compute/runtime/CL/functions/CLPoolingLayer.h"
 #include "arm_compute/runtime/CL/functions/CLSoftmaxLayer.h"
 #include "tests/CL/CLAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/AlexNetFixture.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/CL/SYSTEM/LeNet5.cpp b/tests/benchmark/CL/SYSTEM/LeNet5.cpp
index 35ebc7d8fb..21e5432947 100644
--- a/tests/benchmark/CL/SYSTEM/LeNet5.cpp
+++ b/tests/benchmark/CL/SYSTEM/LeNet5.cpp
@@ -31,10 +31,10 @@
 #include "arm_compute/runtime/CL/functions/CLPoolingLayer.h"
 #include "arm_compute/runtime/CL/functions/CLSoftmaxLayer.h"
 #include "tests/CL/CLAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/LeNet5Fixture.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/NEON/ActivationLayer.cpp b/tests/benchmark/NEON/ActivationLayer.cpp
index 0b699f2501..a00e7ec2e6 100644
--- a/tests/benchmark/NEON/ActivationLayer.cpp
+++ b/tests/benchmark/NEON/ActivationLayer.cpp
@@ -27,7 +27,6 @@
 #include "arm_compute/runtime/Tensor.h"
 #include "arm_compute/runtime/TensorAllocator.h"
 #include "tests/NEON/Accessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/ActivationLayerFixture.h"
 #include "tests/datasets/system_tests/alexnet/AlexNetActivationLayerDataset.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1ActivationLayerDataset.h"
@@ -38,6 +37,7 @@
 #include "tests/datasets/system_tests/yolo/v2/YOLOV2ActivationLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/NEON/BatchNormalizationLayer.cpp b/tests/benchmark/NEON/BatchNormalizationLayer.cpp
index 1b0d484404..5bfd234704 100644
--- a/tests/benchmark/NEON/BatchNormalizationLayer.cpp
+++ b/tests/benchmark/NEON/BatchNormalizationLayer.cpp
@@ -27,9 +27,9 @@
 #include "arm_compute/runtime/Tensor.h"
 #include "arm_compute/runtime/TensorAllocator.h"
 #include "tests/NEON/Accessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 #include "tests/benchmark/fixtures/BatchNormalizationLayerFixture.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4BatchNormalizationLayerDataset.h"
diff --git a/tests/benchmark/NEON/ConvolutionLayer.cpp b/tests/benchmark/NEON/ConvolutionLayer.cpp
index 6d2857d621..effcf63b8b 100644
--- a/tests/benchmark/NEON/ConvolutionLayer.cpp
+++ b/tests/benchmark/NEON/ConvolutionLayer.cpp
@@ -27,7 +27,6 @@
 #include "arm_compute/runtime/Tensor.h"
 #include "arm_compute/runtime/TensorAllocator.h"
 #include "tests/NEON/Accessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/ConvolutionLayerFixture.h"
 #include "tests/datasets/system_tests/alexnet/AlexNetConvolutionLayerDataset.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1ConvolutionLayerDataset.h"
@@ -38,6 +37,7 @@
 #include "tests/datasets/system_tests/yolo/v2/YOLOV2ConvolutionLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/NEON/DirectConvolutionLayer.cpp b/tests/benchmark/NEON/DirectConvolutionLayer.cpp
index a9dd4b94ea..e25a072599 100644
--- a/tests/benchmark/NEON/DirectConvolutionLayer.cpp
+++ b/tests/benchmark/NEON/DirectConvolutionLayer.cpp
@@ -27,7 +27,6 @@
 #include "arm_compute/runtime/Tensor.h"
 #include "arm_compute/runtime/TensorAllocator.h"
 #include "tests/NEON/Accessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/ConvolutionLayerFixture.h"
 #include "tests/datasets/DirectConvolutionLayerDataset.h"
 #include "tests/datasets/system_tests/alexnet/AlexNetConvolutionLayerDataset.h"
@@ -38,6 +37,7 @@
 #include "tests/datasets/system_tests/yolo/v2/YOLOV2ConvolutionLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/NEON/Floor.cpp b/tests/benchmark/NEON/Floor.cpp
index 4bc0c05fdc..824aee95ec 100644
--- a/tests/benchmark/NEON/Floor.cpp
+++ b/tests/benchmark/NEON/Floor.cpp
@@ -27,11 +27,11 @@
 #include "arm_compute/runtime/Tensor.h"
 #include "arm_compute/runtime/TensorAllocator.h"
 #include "tests/NEON/Accessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/FloorFixture.h"
 #include "tests/datasets/ShapeDatasets.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/NEON/FullyConnectedLayer.cpp b/tests/benchmark/NEON/FullyConnectedLayer.cpp
index 88499b1b6f..c12d7cedb9 100644
--- a/tests/benchmark/NEON/FullyConnectedLayer.cpp
+++ b/tests/benchmark/NEON/FullyConnectedLayer.cpp
@@ -27,7 +27,6 @@
 #include "arm_compute/runtime/Tensor.h"
 #include "arm_compute/runtime/TensorAllocator.h"
 #include "tests/NEON/Accessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/FullyConnectedLayerFixture.h"
 #include "tests/datasets/system_tests/alexnet/AlexNetFullyConnectedLayerDataset.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1FullyConnectedLayerDataset.h"
@@ -36,6 +35,7 @@
 #include "tests/datasets/system_tests/vgg/vgg16/VGG16FullyConnectedLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/NEON/GEMM.cpp b/tests/benchmark/NEON/GEMM.cpp
index 2c5a726784..ddc54e633e 100644
--- a/tests/benchmark/NEON/GEMM.cpp
+++ b/tests/benchmark/NEON/GEMM.cpp
@@ -27,12 +27,12 @@
 #include "arm_compute/runtime/Tensor.h"
 #include "arm_compute/runtime/TensorAllocator.h"
 #include "tests/NEON/Accessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/GEMMFixture.h"
 #include "tests/datasets/MatrixMultiplyGEMMDataset.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1GEMMDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/NEON/NormalizationLayer.cpp b/tests/benchmark/NEON/NormalizationLayer.cpp
index 04887951bd..383cec98d0 100644
--- a/tests/benchmark/NEON/NormalizationLayer.cpp
+++ b/tests/benchmark/NEON/NormalizationLayer.cpp
@@ -27,12 +27,12 @@
 #include "arm_compute/runtime/Tensor.h"
 #include "arm_compute/runtime/TensorAllocator.h"
 #include "tests/NEON/Accessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/NormalizationLayerFixture.h"
 #include "tests/datasets/system_tests/alexnet/AlexNetNormalizationLayerDataset.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1NormalizationLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/NEON/PoolingLayer.cpp b/tests/benchmark/NEON/PoolingLayer.cpp
index 05a173b283..8d4e0a086d 100644
--- a/tests/benchmark/NEON/PoolingLayer.cpp
+++ b/tests/benchmark/NEON/PoolingLayer.cpp
@@ -27,7 +27,6 @@
 #include "arm_compute/runtime/Tensor.h"
 #include "arm_compute/runtime/TensorAllocator.h"
 #include "tests/NEON/Accessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/PoolingLayerFixture.h"
 #include "tests/datasets/system_tests/alexnet/AlexNetPoolingLayerDataset.h"
 #include "tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1PoolingLayerDataset.h"
@@ -38,6 +37,7 @@
 #include "tests/datasets/system_tests/yolo/v2/YOLOV2PoolingLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/NEON/ROIPoolingLayer.cpp b/tests/benchmark/NEON/ROIPoolingLayer.cpp
index d3a7e4b3af..64b3570715 100644
--- a/tests/benchmark/NEON/ROIPoolingLayer.cpp
+++ b/tests/benchmark/NEON/ROIPoolingLayer.cpp
@@ -29,11 +29,11 @@
 #include "arm_compute/runtime/TensorAllocator.h"
 #include "tests/NEON/Accessor.h"
 #include "tests/NEON/ArrayAccessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/ROIPoolingLayerFixture.h"
 #include "tests/datasets/ROIPoolingLayerDataset.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/NEON/SYSTEM/AlexNet.cpp b/tests/benchmark/NEON/SYSTEM/AlexNet.cpp
index cd48e5db25..19598e3ca5 100644
--- a/tests/benchmark/NEON/SYSTEM/AlexNet.cpp
+++ b/tests/benchmark/NEON/SYSTEM/AlexNet.cpp
@@ -34,10 +34,10 @@
 #include "arm_compute/runtime/Tensor.h"
 #include "arm_compute/runtime/TensorAllocator.h"
 #include "tests/NEON/Accessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/AlexNetFixture.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/benchmark/NEON/SYSTEM/LeNet5.cpp b/tests/benchmark/NEON/SYSTEM/LeNet5.cpp
index 46ccfc01d5..5075da50cf 100644
--- a/tests/benchmark/NEON/SYSTEM/LeNet5.cpp
+++ b/tests/benchmark/NEON/SYSTEM/LeNet5.cpp
@@ -31,10 +31,10 @@
 #include "arm_compute/runtime/Tensor.h"
 #include "arm_compute/runtime/TensorAllocator.h"
 #include "tests/NEON/Accessor.h"
-#include "tests/TypePrinter.h"
 #include "tests/benchmark/fixtures/LeNet5Fixture.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/datasets/BatchNormalizationLayerDataset.h b/tests/datasets/BatchNormalizationLayerDataset.h
index 25e65d931c..fceff1edc7 100644
--- a/tests/datasets/BatchNormalizationLayerDataset.h
+++ b/tests/datasets/BatchNormalizationLayerDataset.h
@@ -24,7 +24,7 @@
 #ifndef ARM_COMPUTE_TEST_BATCHNORMALIZATION_LAYER_DATASET
 #define ARM_COMPUTE_TEST_BATCHNORMALIZATION_LAYER_DATASET
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/BorderModeDataset.h b/tests/datasets/BorderModeDataset.h
index df024a7dbf..3b8a08ce61 100644
--- a/tests/datasets/BorderModeDataset.h
+++ b/tests/datasets/BorderModeDataset.h
@@ -27,6 +27,13 @@
 #include "arm_compute/core/Types.h"
 #include "tests/framework/datasets/ContainerDataset.h"
 
+#include "utils/TypePrinter.h"
+
+#include "arm_compute/core/TensorShape.h"
+#include "arm_compute/core/Types.h"
+
+#include "arm_compute/core/TensorShape.h"
+#include "arm_compute/core/Types.h"
 #include <vector>
 
 namespace arm_compute
diff --git a/tests/datasets/ConvolutionLayerDataset.h b/tests/datasets/ConvolutionLayerDataset.h
index ba11bd5d6d..6e2d2a142e 100644
--- a/tests/datasets/ConvolutionLayerDataset.h
+++ b/tests/datasets/ConvolutionLayerDataset.h
@@ -24,7 +24,7 @@
 #ifndef ARM_COMPUTE_TEST_CONVOLUTION_LAYER_DATASET
 #define ARM_COMPUTE_TEST_CONVOLUTION_LAYER_DATASET
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/DepthwiseConvolutionDataset.h b/tests/datasets/DepthwiseConvolutionDataset.h
index bdc949501e..6a7af63d05 100644
--- a/tests/datasets/DepthwiseConvolutionDataset.h
+++ b/tests/datasets/DepthwiseConvolutionDataset.h
@@ -24,7 +24,7 @@
 #ifndef ARM_COMPUTE_TEST_DEPTHWISE_CONVOLUTION_DATASET
 #define ARM_COMPUTE_TEST_DEPTHWISE_CONVOLUTION_DATASET
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/DepthwiseSeparableConvolutionLayerDataset.h b/tests/datasets/DepthwiseSeparableConvolutionLayerDataset.h
index 4391379833..6b39d3a5c8 100644
--- a/tests/datasets/DepthwiseSeparableConvolutionLayerDataset.h
+++ b/tests/datasets/DepthwiseSeparableConvolutionLayerDataset.h
@@ -24,7 +24,7 @@
 #ifndef ARM_COMPUTE_TEST_DEPTHWISE_SEPARABLE_CONVOLUTION_LAYER_DATASET
 #define ARM_COMPUTE_TEST_DEPTHWISE_SEPARABLE_CONVOLUTION_LAYER_DATASET
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/DirectConvolutionLayerDataset.h b/tests/datasets/DirectConvolutionLayerDataset.h
index dca38b5858..294f44fa42 100644
--- a/tests/datasets/DirectConvolutionLayerDataset.h
+++ b/tests/datasets/DirectConvolutionLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/ConvolutionLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/FullyConnectedLayerDataset.h b/tests/datasets/FullyConnectedLayerDataset.h
index 8401e39ece..5789954e85 100644
--- a/tests/datasets/FullyConnectedLayerDataset.h
+++ b/tests/datasets/FullyConnectedLayerDataset.h
@@ -24,7 +24,7 @@
 #ifndef ARM_COMPUTE_TEST_FULLYCONNECTED_LAYER_DATASET
 #define ARM_COMPUTE_TEST_FULLYCONNECTED_LAYER_DATASET
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/GEMMDataset.h b/tests/datasets/GEMMDataset.h
index bb8a328b0c..9491582641 100644
--- a/tests/datasets/GEMMDataset.h
+++ b/tests/datasets/GEMMDataset.h
@@ -24,7 +24,7 @@
 #ifndef ARM_COMPUTE_TEST_GEMM_DATASET
 #define ARM_COMPUTE_TEST_GEMM_DATASET
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 
diff --git a/tests/datasets/LargeConvolutionLayerDataset.h b/tests/datasets/LargeConvolutionLayerDataset.h
index 721530ae19..086b2e3def 100644
--- a/tests/datasets/LargeConvolutionLayerDataset.h
+++ b/tests/datasets/LargeConvolutionLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/ConvolutionLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/LargeGEMMDataset.h b/tests/datasets/LargeGEMMDataset.h
index 37cdfaa99f..638eb904ad 100644
--- a/tests/datasets/LargeGEMMDataset.h
+++ b/tests/datasets/LargeGEMMDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/GEMMDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/MatrixMultiplyGEMMDataset.h b/tests/datasets/MatrixMultiplyGEMMDataset.h
index 771403ba96..718c32b466 100644
--- a/tests/datasets/MatrixMultiplyGEMMDataset.h
+++ b/tests/datasets/MatrixMultiplyGEMMDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/GEMMDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 
diff --git a/tests/datasets/MatrixPatternDataset.h b/tests/datasets/MatrixPatternDataset.h
index 4c1414996d..423aeb8627 100644
--- a/tests/datasets/MatrixPatternDataset.h
+++ b/tests/datasets/MatrixPatternDataset.h
@@ -27,6 +27,13 @@
 #include "arm_compute/core/Types.h"
 #include "tests/framework/datasets/ContainerDataset.h"
 
+#include "utils/TypePrinter.h"
+
+#include "arm_compute/core/TensorShape.h"
+#include "arm_compute/core/Types.h"
+
+#include "arm_compute/core/TensorShape.h"
+#include "arm_compute/core/Types.h"
 #include <vector>
 
 namespace arm_compute
diff --git a/tests/datasets/MobileNetDepthwiseConvolutionDataset.h b/tests/datasets/MobileNetDepthwiseConvolutionDataset.h
index c417f18a51..c9d98d4185 100644
--- a/tests/datasets/MobileNetDepthwiseConvolutionDataset.h
+++ b/tests/datasets/MobileNetDepthwiseConvolutionDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/DepthwiseConvolutionDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/MobileNetDepthwiseSeparableConvolutionLayerDataset.h b/tests/datasets/MobileNetDepthwiseSeparableConvolutionLayerDataset.h
index c7784c399c..f6137ee74c 100644
--- a/tests/datasets/MobileNetDepthwiseSeparableConvolutionLayerDataset.h
+++ b/tests/datasets/MobileNetDepthwiseSeparableConvolutionLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/DepthwiseSeparableConvolutionLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/PoolingLayerDataset.h b/tests/datasets/PoolingLayerDataset.h
index e5e9cd70f0..56ec3b87d8 100644
--- a/tests/datasets/PoolingLayerDataset.h
+++ b/tests/datasets/PoolingLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 namespace arm_compute
 {
diff --git a/tests/datasets/ROIPoolingLayerDataset.h b/tests/datasets/ROIPoolingLayerDataset.h
index 65d589eb97..a1879d3b54 100644
--- a/tests/datasets/ROIPoolingLayerDataset.h
+++ b/tests/datasets/ROIPoolingLayerDataset.h
@@ -24,7 +24,7 @@
 #ifndef ARM_COMPUTE_TEST_ROI_POOLING_LAYER_DATASET
 #define ARM_COMPUTE_TEST_ROI_POOLING_LAYER_DATASET
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/RandomBatchNormalizationLayerDataset.h b/tests/datasets/RandomBatchNormalizationLayerDataset.h
index f4c61e06d5..10908871b6 100644
--- a/tests/datasets/RandomBatchNormalizationLayerDataset.h
+++ b/tests/datasets/RandomBatchNormalizationLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/BatchNormalizationLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/ReductionOperationDataset.h b/tests/datasets/ReductionOperationDataset.h
index ce1bcb8501..e710588921 100644
--- a/tests/datasets/ReductionOperationDataset.h
+++ b/tests/datasets/ReductionOperationDataset.h
@@ -25,8 +25,8 @@
 #define __ARM_COMPUTE_TEST_REDUCTION_OPERATION_DATASET_H__
 
 #include "arm_compute/core/Types.h"
-#include "tests/TypePrinter.h"
 #include "tests/framework/datasets/ContainerDataset.h"
+#include "utils/TypePrinter.h"
 
 #include <vector>
 
diff --git a/tests/datasets/ReshapeLayerDataset.h b/tests/datasets/ReshapeLayerDataset.h
index db7b0b3e57..fa938a63ed 100644
--- a/tests/datasets/ReshapeLayerDataset.h
+++ b/tests/datasets/ReshapeLayerDataset.h
@@ -24,7 +24,7 @@
 #ifndef ARM_COMPUTE_TEST_RESHAPE_LAYER_DATASET
 #define ARM_COMPUTE_TEST_RESHAPE_LAYER_DATASET
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 
diff --git a/tests/datasets/SmallConvolutionLayerDataset.h b/tests/datasets/SmallConvolutionLayerDataset.h
index 1a26fa5917..8eda2e87fe 100644
--- a/tests/datasets/SmallConvolutionLayerDataset.h
+++ b/tests/datasets/SmallConvolutionLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/ConvolutionLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/SmallGEMMDataset.h b/tests/datasets/SmallGEMMDataset.h
index 0cc3c3aa8c..5d59c1d3b3 100644
--- a/tests/datasets/SmallGEMMDataset.h
+++ b/tests/datasets/SmallGEMMDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/GEMMDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/alexnet/AlexNetActivationLayerDataset.h b/tests/datasets/system_tests/alexnet/AlexNetActivationLayerDataset.h
index 4030e97436..74c5fda235 100644
--- a/tests/datasets/system_tests/alexnet/AlexNetActivationLayerDataset.h
+++ b/tests/datasets/system_tests/alexnet/AlexNetActivationLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/framework/datasets/Datasets.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/alexnet/AlexNetConvolutionLayerDataset.h b/tests/datasets/system_tests/alexnet/AlexNetConvolutionLayerDataset.h
index d0b901a9a1..dcf96c7991 100644
--- a/tests/datasets/system_tests/alexnet/AlexNetConvolutionLayerDataset.h
+++ b/tests/datasets/system_tests/alexnet/AlexNetConvolutionLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/ConvolutionLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/alexnet/AlexNetFullyConnectedLayerDataset.h b/tests/datasets/system_tests/alexnet/AlexNetFullyConnectedLayerDataset.h
index 50b0f7d75d..3d43e6f7db 100644
--- a/tests/datasets/system_tests/alexnet/AlexNetFullyConnectedLayerDataset.h
+++ b/tests/datasets/system_tests/alexnet/AlexNetFullyConnectedLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/FullyConnectedLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/alexnet/AlexNetNormalizationLayerDataset.h b/tests/datasets/system_tests/alexnet/AlexNetNormalizationLayerDataset.h
index 33b7423442..6f68fd7722 100644
--- a/tests/datasets/system_tests/alexnet/AlexNetNormalizationLayerDataset.h
+++ b/tests/datasets/system_tests/alexnet/AlexNetNormalizationLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/framework/datasets/Datasets.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/alexnet/AlexNetPoolingLayerDataset.h b/tests/datasets/system_tests/alexnet/AlexNetPoolingLayerDataset.h
index ab2749b36e..739d24ca18 100644
--- a/tests/datasets/system_tests/alexnet/AlexNetPoolingLayerDataset.h
+++ b/tests/datasets/system_tests/alexnet/AlexNetPoolingLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/PoolingLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1ActivationLayerDataset.h b/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1ActivationLayerDataset.h
index dc4ffe4e4e..76087fa3d5 100644
--- a/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1ActivationLayerDataset.h
+++ b/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1ActivationLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/framework/datasets/Datasets.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1ConvolutionLayerDataset.h b/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1ConvolutionLayerDataset.h
index a4002d124d..1415ff1daa 100644
--- a/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1ConvolutionLayerDataset.h
+++ b/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1ConvolutionLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/ConvolutionLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1FullyConnectedLayerDataset.h b/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1FullyConnectedLayerDataset.h
index 80a3473bc7..cde9ae6179 100644
--- a/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1FullyConnectedLayerDataset.h
+++ b/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1FullyConnectedLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/FullyConnectedLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1GEMMDataset.h b/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1GEMMDataset.h
index 806ca9378a..3b5c9acc8a 100644
--- a/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1GEMMDataset.h
+++ b/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1GEMMDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/GEMMDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 
diff --git a/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1NormalizationLayerDataset.h b/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1NormalizationLayerDataset.h
index d8fd061b66..eb6903faef 100644
--- a/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1NormalizationLayerDataset.h
+++ b/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1NormalizationLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/framework/datasets/Datasets.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1PoolingLayerDataset.h b/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1PoolingLayerDataset.h
index 6164bbacea..b775667ac8 100644
--- a/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1PoolingLayerDataset.h
+++ b/tests/datasets/system_tests/googlenet/inceptionv1/GoogLeNetInceptionV1PoolingLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/PoolingLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4ActivationLayerDataset.h b/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4ActivationLayerDataset.h
index 2d586390ec..395b1f9866 100644
--- a/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4ActivationLayerDataset.h
+++ b/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4ActivationLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/framework/datasets/Datasets.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4BatchNormalizationLayerDataset.h b/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4BatchNormalizationLayerDataset.h
index d96410f9eb..3a22c7807f 100644
--- a/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4BatchNormalizationLayerDataset.h
+++ b/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4BatchNormalizationLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/BatchNormalizationLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4ConvolutionLayerDataset.h b/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4ConvolutionLayerDataset.h
index a050d3da74..3dd8a7c348 100644
--- a/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4ConvolutionLayerDataset.h
+++ b/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4ConvolutionLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/ConvolutionLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4FullyConnectedLayerDataset.h b/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4FullyConnectedLayerDataset.h
index 8cf59e8207..5a0af2b6c5 100644
--- a/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4FullyConnectedLayerDataset.h
+++ b/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4FullyConnectedLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/FullyConnectedLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4PoolingLayerDataset.h b/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4PoolingLayerDataset.h
index 3b1ee8dcb4..a58fd797ec 100644
--- a/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4PoolingLayerDataset.h
+++ b/tests/datasets/system_tests/googlenet/inceptionv4/GoogLeNetInceptionV4PoolingLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/PoolingLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/lenet5/LeNet5ActivationLayerDataset.h b/tests/datasets/system_tests/lenet5/LeNet5ActivationLayerDataset.h
index 4c88cde58d..021bfee18b 100644
--- a/tests/datasets/system_tests/lenet5/LeNet5ActivationLayerDataset.h
+++ b/tests/datasets/system_tests/lenet5/LeNet5ActivationLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/framework/datasets/Datasets.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/lenet5/LeNet5ConvolutionLayerDataset.h b/tests/datasets/system_tests/lenet5/LeNet5ConvolutionLayerDataset.h
index e1d3ead65b..3f28627b08 100644
--- a/tests/datasets/system_tests/lenet5/LeNet5ConvolutionLayerDataset.h
+++ b/tests/datasets/system_tests/lenet5/LeNet5ConvolutionLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/ConvolutionLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/lenet5/LeNet5FullyConnectedLayerDataset.h b/tests/datasets/system_tests/lenet5/LeNet5FullyConnectedLayerDataset.h
index 343ecc0190..74e8d2ce7b 100644
--- a/tests/datasets/system_tests/lenet5/LeNet5FullyConnectedLayerDataset.h
+++ b/tests/datasets/system_tests/lenet5/LeNet5FullyConnectedLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/FullyConnectedLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/lenet5/LeNet5PoolingLayerDataset.h b/tests/datasets/system_tests/lenet5/LeNet5PoolingLayerDataset.h
index bc2de7b235..bcd11dc5b9 100644
--- a/tests/datasets/system_tests/lenet5/LeNet5PoolingLayerDataset.h
+++ b/tests/datasets/system_tests/lenet5/LeNet5PoolingLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/PoolingLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDataset.h b/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDataset.h
index 7644387384..7f4bf4d549 100644
--- a/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDataset.h
+++ b/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/framework/datasets/Datasets.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDataset.h b/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDataset.h
index 2d447b0156..2bb9db7240 100644
--- a/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDataset.h
+++ b/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/ConvolutionLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h b/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h
index 69d04db7e1..cb2edbc5ec 100644
--- a/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h
+++ b/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/PoolingLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/vgg/vgg16/VGG16ActivationLayerDataset.h b/tests/datasets/system_tests/vgg/vgg16/VGG16ActivationLayerDataset.h
index 66301ddf08..93de996b6c 100644
--- a/tests/datasets/system_tests/vgg/vgg16/VGG16ActivationLayerDataset.h
+++ b/tests/datasets/system_tests/vgg/vgg16/VGG16ActivationLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/framework/datasets/Datasets.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/vgg/vgg16/VGG16ConvolutionLayerDataset.h b/tests/datasets/system_tests/vgg/vgg16/VGG16ConvolutionLayerDataset.h
index 36cb5d93aa..fe31da287c 100644
--- a/tests/datasets/system_tests/vgg/vgg16/VGG16ConvolutionLayerDataset.h
+++ b/tests/datasets/system_tests/vgg/vgg16/VGG16ConvolutionLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/ConvolutionLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/vgg/vgg16/VGG16FullyConnectedLayerDataset.h b/tests/datasets/system_tests/vgg/vgg16/VGG16FullyConnectedLayerDataset.h
index d34dc3f87e..3b40b34653 100644
--- a/tests/datasets/system_tests/vgg/vgg16/VGG16FullyConnectedLayerDataset.h
+++ b/tests/datasets/system_tests/vgg/vgg16/VGG16FullyConnectedLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/FullyConnectedLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/vgg/vgg16/VGG16PoolingLayerDataset.h b/tests/datasets/system_tests/vgg/vgg16/VGG16PoolingLayerDataset.h
index 4db20870f2..affbcc5022 100644
--- a/tests/datasets/system_tests/vgg/vgg16/VGG16PoolingLayerDataset.h
+++ b/tests/datasets/system_tests/vgg/vgg16/VGG16PoolingLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/PoolingLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/yolo/v2/YOLOV2ActivationLayerDataset.h b/tests/datasets/system_tests/yolo/v2/YOLOV2ActivationLayerDataset.h
index 8a2cd1665e..f30069e55d 100644
--- a/tests/datasets/system_tests/yolo/v2/YOLOV2ActivationLayerDataset.h
+++ b/tests/datasets/system_tests/yolo/v2/YOLOV2ActivationLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/framework/datasets/Datasets.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/yolo/v2/YOLOV2BatchNormalizationLayerDataset.h b/tests/datasets/system_tests/yolo/v2/YOLOV2BatchNormalizationLayerDataset.h
index cc6a6dcccc..4d2cba8b4b 100644
--- a/tests/datasets/system_tests/yolo/v2/YOLOV2BatchNormalizationLayerDataset.h
+++ b/tests/datasets/system_tests/yolo/v2/YOLOV2BatchNormalizationLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/BatchNormalizationLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/yolo/v2/YOLOV2ConvolutionLayerDataset.h b/tests/datasets/system_tests/yolo/v2/YOLOV2ConvolutionLayerDataset.h
index 41f48b0958..b4cb3577e8 100644
--- a/tests/datasets/system_tests/yolo/v2/YOLOV2ConvolutionLayerDataset.h
+++ b/tests/datasets/system_tests/yolo/v2/YOLOV2ConvolutionLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/ConvolutionLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/datasets/system_tests/yolo/v2/YOLOV2PoolingLayerDataset.h b/tests/datasets/system_tests/yolo/v2/YOLOV2PoolingLayerDataset.h
index 3763d63588..62222f3a88 100644
--- a/tests/datasets/system_tests/yolo/v2/YOLOV2PoolingLayerDataset.h
+++ b/tests/datasets/system_tests/yolo/v2/YOLOV2PoolingLayerDataset.h
@@ -26,7 +26,7 @@
 
 #include "tests/datasets/PoolingLayerDataset.h"
 
-#include "tests/TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation/Validation.h b/tests/validation/Validation.h
index 7d4ce57250..38cbadaa9b 100644
--- a/tests/validation/Validation.h
+++ b/tests/validation/Validation.h
@@ -28,10 +28,10 @@
 #include "arm_compute/core/Types.h"
 #include "tests/IAccessor.h"
 #include "tests/SimpleTensor.h"
-#include "tests/TypePrinter.h"
 #include "tests/Utils.h"
 #include "tests/framework/Asserts.h"
 #include "tests/framework/Exceptions.h"
+#include "utils/TypePrinter.h"
 
 #include <iomanip>
 #include <ios>
diff --git a/tests/validation_old/AssetsLibrary.cpp b/tests/validation_old/AssetsLibrary.cpp
index fa8569e458..997af27a9f 100644
--- a/tests/validation_old/AssetsLibrary.cpp
+++ b/tests/validation_old/AssetsLibrary.cpp
@@ -24,8 +24,8 @@
 #include "tests/AssetsLibrary.h"
 
 #include "arm_compute/core/ITensor.h"
-#include "tests/TypePrinter.h"
 #include "tests/Utils.h"
+#include "utils/TypePrinter.h"
 
 #include <cctype>
 #include <fstream>
diff --git a/tests/validation_old/CL/FillBorder.cpp b/tests/validation_old/CL/FillBorder.cpp
index ed47a1eeb3..deb40a1a5a 100644
--- a/tests/validation_old/CL/FillBorder.cpp
+++ b/tests/validation_old/CL/FillBorder.cpp
@@ -22,12 +22,12 @@
  * SOFTWARE.
  */
 #include "CL/CLAccessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/CL/kernels/CLFillBorderKernel.h"
 #include "arm_compute/core/Helpers.h"
diff --git a/tests/validation_old/CL/FixedPoint/FixedPoint_QS8.cpp b/tests/validation_old/CL/FixedPoint/FixedPoint_QS8.cpp
index 3721fb51d7..4f7764c49b 100644
--- a/tests/validation_old/CL/FixedPoint/FixedPoint_QS8.cpp
+++ b/tests/validation_old/CL/FixedPoint/FixedPoint_QS8.cpp
@@ -22,13 +22,13 @@
  * SOFTWARE.
  */
 #include "CL/CLAccessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/CL/CLKernelLibrary.h"
 #include "arm_compute/core/CL/ICLKernel.h"
diff --git a/tests/validation_old/CL/Gaussian3x3.cpp b/tests/validation_old/CL/Gaussian3x3.cpp
index 27f4833289..6aa2d0349e 100644
--- a/tests/validation_old/CL/Gaussian3x3.cpp
+++ b/tests/validation_old/CL/Gaussian3x3.cpp
@@ -23,7 +23,6 @@
  */
 #include "CL/CLAccessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -31,6 +30,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/CL/Gaussian5x5.cpp b/tests/validation_old/CL/Gaussian5x5.cpp
index c187426f4c..32a33d2661 100644
--- a/tests/validation_old/CL/Gaussian5x5.cpp
+++ b/tests/validation_old/CL/Gaussian5x5.cpp
@@ -23,7 +23,6 @@
  */
 #include "CL/CLAccessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -31,6 +30,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/CL/HarrisCorners.cpp b/tests/validation_old/CL/HarrisCorners.cpp
index 2c73679058..ff39918a43 100644
--- a/tests/validation_old/CL/HarrisCorners.cpp
+++ b/tests/validation_old/CL/HarrisCorners.cpp
@@ -22,7 +22,6 @@
  * SOFTWARE.
  */
 #include "CL/CLAccessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -30,6 +29,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/CL/IntegralImage.cpp b/tests/validation_old/CL/IntegralImage.cpp
index ea15b90b2a..7db2b688d6 100644
--- a/tests/validation_old/CL/IntegralImage.cpp
+++ b/tests/validation_old/CL/IntegralImage.cpp
@@ -24,13 +24,13 @@
 
 #include "CL/CLAccessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/CL/MinMaxLocation.cpp b/tests/validation_old/CL/MinMaxLocation.cpp
index 8824215223..06e9388abc 100644
--- a/tests/validation_old/CL/MinMaxLocation.cpp
+++ b/tests/validation_old/CL/MinMaxLocation.cpp
@@ -23,13 +23,13 @@
  */
 
 #include "CL/CLAccessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/CL/PixelWiseMultiplication.cpp b/tests/validation_old/CL/PixelWiseMultiplication.cpp
index f003298a23..5bc71820d0 100644
--- a/tests/validation_old/CL/PixelWiseMultiplication.cpp
+++ b/tests/validation_old/CL/PixelWiseMultiplication.cpp
@@ -22,12 +22,12 @@
  * SOFTWARE.
  */
 #include "CL/CLAccessor.h"
-#include "TypePrinter.h"
 #include "tests/Globals.h"
 #include "tests/Utils.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/runtime/CL/functions/CLPixelWiseMultiplication.h"
 
diff --git a/tests/validation_old/CL/ROIPoolingLayer.cpp b/tests/validation_old/CL/ROIPoolingLayer.cpp
index edd1cccf2a..ec804ccb84 100644
--- a/tests/validation_old/CL/ROIPoolingLayer.cpp
+++ b/tests/validation_old/CL/ROIPoolingLayer.cpp
@@ -23,7 +23,6 @@
  */
 #include "CL/CLAccessor.h"
 #include "CL/CLArrayAccessor.h"
-#include "TypePrinter.h"
 #include "arm_compute/runtime/CL/CLArray.h"
 #include "arm_compute/runtime/CL/functions/CLROIPoolingLayer.h"
 #include "tests/Globals.h"
@@ -32,6 +31,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include <random>
 #include <vector>
diff --git a/tests/validation_old/CL/Sobel3x3.cpp b/tests/validation_old/CL/Sobel3x3.cpp
index a4c779cd5c..cb4e318736 100644
--- a/tests/validation_old/CL/Sobel3x3.cpp
+++ b/tests/validation_old/CL/Sobel3x3.cpp
@@ -24,7 +24,6 @@
 
 #include "CL/CLAccessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -32,6 +31,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/CL/Sobel5x5.cpp b/tests/validation_old/CL/Sobel5x5.cpp
index 7e5dec1209..d035027218 100644
--- a/tests/validation_old/CL/Sobel5x5.cpp
+++ b/tests/validation_old/CL/Sobel5x5.cpp
@@ -24,7 +24,6 @@
 
 #include "CL/CLAccessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -32,6 +31,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/CL/Threshold.cpp b/tests/validation_old/CL/Threshold.cpp
index 74ddd6873e..185790782d 100644
--- a/tests/validation_old/CL/Threshold.cpp
+++ b/tests/validation_old/CL/Threshold.cpp
@@ -23,7 +23,6 @@
  */
 #include "CL/CLAccessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -31,6 +30,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/dataset/ThresholdDataset.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/CL/WarpPerspective.cpp b/tests/validation_old/CL/WarpPerspective.cpp
index 6252361003..3d3da7a23e 100644
--- a/tests/validation_old/CL/WarpPerspective.cpp
+++ b/tests/validation_old/CL/WarpPerspective.cpp
@@ -23,7 +23,6 @@
  */
 #include "CL/CLAccessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -32,6 +31,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/AbsoluteDifference.cpp b/tests/validation_old/NEON/AbsoluteDifference.cpp
index aa866fff49..cde0b46268 100644
--- a/tests/validation_old/NEON/AbsoluteDifference.cpp
+++ b/tests/validation_old/NEON/AbsoluteDifference.cpp
@@ -23,13 +23,13 @@
  */
 #include "NEON/Accessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/Accumulate.cpp b/tests/validation_old/NEON/Accumulate.cpp
index eb680a383d..a173338542 100644
--- a/tests/validation_old/NEON/Accumulate.cpp
+++ b/tests/validation_old/NEON/Accumulate.cpp
@@ -23,13 +23,13 @@
  */
 #include "NEON/Accessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/AccumulateSquared.cpp b/tests/validation_old/NEON/AccumulateSquared.cpp
index 29b5edf41b..f4eea7d764 100644
--- a/tests/validation_old/NEON/AccumulateSquared.cpp
+++ b/tests/validation_old/NEON/AccumulateSquared.cpp
@@ -23,13 +23,13 @@
  */
 #include "NEON/Accessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/AccumulateWeighted.cpp b/tests/validation_old/NEON/AccumulateWeighted.cpp
index c59c1edbc8..758e437338 100644
--- a/tests/validation_old/NEON/AccumulateWeighted.cpp
+++ b/tests/validation_old/NEON/AccumulateWeighted.cpp
@@ -23,13 +23,13 @@
  */
 #include "NEON/Accessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/FillBorder.cpp b/tests/validation_old/NEON/FillBorder.cpp
index 277bbf2c0a..cbf637e7f8 100644
--- a/tests/validation_old/NEON/FillBorder.cpp
+++ b/tests/validation_old/NEON/FillBorder.cpp
@@ -22,12 +22,12 @@
  * SOFTWARE.
  */
 #include "NEON/Accessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/NEON/kernels/NEFillBorderKernel.h"
diff --git a/tests/validation_old/NEON/Fixedpoint/Exp_QS16.cpp b/tests/validation_old/NEON/Fixedpoint/Exp_QS16.cpp
index 66115879aa..a0e785e112 100644
--- a/tests/validation_old/NEON/Fixedpoint/Exp_QS16.cpp
+++ b/tests/validation_old/NEON/Fixedpoint/Exp_QS16.cpp
@@ -22,13 +22,13 @@
  * SOFTWARE.
  */
 #include "NEON/Accessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/ReferenceCPP.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/NEON/NEFixedPoint.h"
diff --git a/tests/validation_old/NEON/Fixedpoint/Exp_QS8.cpp b/tests/validation_old/NEON/Fixedpoint/Exp_QS8.cpp
index 9e8096fa75..dbae11640a 100644
--- a/tests/validation_old/NEON/Fixedpoint/Exp_QS8.cpp
+++ b/tests/validation_old/NEON/Fixedpoint/Exp_QS8.cpp
@@ -22,13 +22,13 @@
  * SOFTWARE.
  */
 #include "NEON/Accessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/ReferenceCPP.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/NEON/NEFixedPoint.h"
diff --git a/tests/validation_old/NEON/Fixedpoint/Invsqrt_QS16.cpp b/tests/validation_old/NEON/Fixedpoint/Invsqrt_QS16.cpp
index f56707a93d..1d06e0799d 100644
--- a/tests/validation_old/NEON/Fixedpoint/Invsqrt_QS16.cpp
+++ b/tests/validation_old/NEON/Fixedpoint/Invsqrt_QS16.cpp
@@ -22,13 +22,13 @@
  * SOFTWARE.
  */
 #include "NEON/Accessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/ReferenceCPP.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/NEON/NEFixedPoint.h"
diff --git a/tests/validation_old/NEON/Fixedpoint/Invsqrt_QS8.cpp b/tests/validation_old/NEON/Fixedpoint/Invsqrt_QS8.cpp
index fb33fd4632..ea21b8f4d6 100644
--- a/tests/validation_old/NEON/Fixedpoint/Invsqrt_QS8.cpp
+++ b/tests/validation_old/NEON/Fixedpoint/Invsqrt_QS8.cpp
@@ -22,13 +22,13 @@
  * SOFTWARE.
  */
 #include "NEON/Accessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/ReferenceCPP.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/NEON/NEFixedPoint.h"
diff --git a/tests/validation_old/NEON/Fixedpoint/Log_QS16.cpp b/tests/validation_old/NEON/Fixedpoint/Log_QS16.cpp
index 6485b2031c..d85112abcc 100644
--- a/tests/validation_old/NEON/Fixedpoint/Log_QS16.cpp
+++ b/tests/validation_old/NEON/Fixedpoint/Log_QS16.cpp
@@ -22,13 +22,13 @@
  * SOFTWARE.
  */
 #include "NEON/Accessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/ReferenceCPP.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/NEON/NEFixedPoint.h"
diff --git a/tests/validation_old/NEON/Fixedpoint/Log_QS8.cpp b/tests/validation_old/NEON/Fixedpoint/Log_QS8.cpp
index 21012c52b0..41b285650a 100644
--- a/tests/validation_old/NEON/Fixedpoint/Log_QS8.cpp
+++ b/tests/validation_old/NEON/Fixedpoint/Log_QS8.cpp
@@ -22,13 +22,13 @@
  * SOFTWARE.
  */
 #include "NEON/Accessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/ReferenceCPP.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/NEON/NEFixedPoint.h"
diff --git a/tests/validation_old/NEON/Fixedpoint/Reciprocal_QS16.cpp b/tests/validation_old/NEON/Fixedpoint/Reciprocal_QS16.cpp
index 5630a3391a..9605595d80 100644
--- a/tests/validation_old/NEON/Fixedpoint/Reciprocal_QS16.cpp
+++ b/tests/validation_old/NEON/Fixedpoint/Reciprocal_QS16.cpp
@@ -22,13 +22,13 @@
  * SOFTWARE.
  */
 #include "NEON/Accessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/ReferenceCPP.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/NEON/NEFixedPoint.h"
diff --git a/tests/validation_old/NEON/Fixedpoint/Reciprocal_QS8.cpp b/tests/validation_old/NEON/Fixedpoint/Reciprocal_QS8.cpp
index 23f98acc40..7a1c5b72a7 100644
--- a/tests/validation_old/NEON/Fixedpoint/Reciprocal_QS8.cpp
+++ b/tests/validation_old/NEON/Fixedpoint/Reciprocal_QS8.cpp
@@ -22,13 +22,13 @@
  * SOFTWARE.
  */
 #include "NEON/Accessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/ReferenceCPP.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/NEON/NEFixedPoint.h"
diff --git a/tests/validation_old/NEON/Gaussian3x3.cpp b/tests/validation_old/NEON/Gaussian3x3.cpp
index becd9196ea..b1fdc5e069 100644
--- a/tests/validation_old/NEON/Gaussian3x3.cpp
+++ b/tests/validation_old/NEON/Gaussian3x3.cpp
@@ -23,7 +23,6 @@
  */
 #include "NEON/Accessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -31,6 +30,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/Gaussian5x5.cpp b/tests/validation_old/NEON/Gaussian5x5.cpp
index 240285afb6..c98d21b35c 100644
--- a/tests/validation_old/NEON/Gaussian5x5.cpp
+++ b/tests/validation_old/NEON/Gaussian5x5.cpp
@@ -23,7 +23,6 @@
  */
 #include "NEON/Accessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -31,6 +30,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/HarrisCorners.cpp b/tests/validation_old/NEON/HarrisCorners.cpp
index 809e61c053..172be87dc2 100644
--- a/tests/validation_old/NEON/HarrisCorners.cpp
+++ b/tests/validation_old/NEON/HarrisCorners.cpp
@@ -23,7 +23,6 @@
  */
 #include "NEON/Accessor.h"
 #include "NEON/Helper.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -31,6 +30,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/IntegralImage.cpp b/tests/validation_old/NEON/IntegralImage.cpp
index 69654b2585..f42393a745 100644
--- a/tests/validation_old/NEON/IntegralImage.cpp
+++ b/tests/validation_old/NEON/IntegralImage.cpp
@@ -24,13 +24,13 @@
 
 #include "NEON/Accessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/MinMaxLocation.cpp b/tests/validation_old/NEON/MinMaxLocation.cpp
index c41745a636..1b258d19d6 100644
--- a/tests/validation_old/NEON/MinMaxLocation.cpp
+++ b/tests/validation_old/NEON/MinMaxLocation.cpp
@@ -25,13 +25,13 @@
 #include "NEON/Accessor.h"
 #include "NEON/Helper.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/PixelWiseMultiplication.cpp b/tests/validation_old/NEON/PixelWiseMultiplication.cpp
index 60eb82ef84..f352fc4390 100644
--- a/tests/validation_old/NEON/PixelWiseMultiplication.cpp
+++ b/tests/validation_old/NEON/PixelWiseMultiplication.cpp
@@ -23,13 +23,13 @@
  */
 #include "NEON/Accessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
 #include "tests/validation_old/Datasets.h"
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/ROIPoolingLayer.cpp b/tests/validation_old/NEON/ROIPoolingLayer.cpp
index 2046beb196..a8d5637412 100644
--- a/tests/validation_old/NEON/ROIPoolingLayer.cpp
+++ b/tests/validation_old/NEON/ROIPoolingLayer.cpp
@@ -23,7 +23,6 @@
  */
 #include "NEON/Accessor.h"
 #include "NEON/ArrayAccessor.h"
-#include "TypePrinter.h"
 #include "arm_compute/runtime/NEON/functions/NEROIPoolingLayer.h"
 #include "tests/Globals.h"
 #include "tests/Utils.h"
@@ -31,6 +30,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include <random>
 #include <vector>
diff --git a/tests/validation_old/NEON/Sobel3x3.cpp b/tests/validation_old/NEON/Sobel3x3.cpp
index cb249e1a58..302364048a 100644
--- a/tests/validation_old/NEON/Sobel3x3.cpp
+++ b/tests/validation_old/NEON/Sobel3x3.cpp
@@ -22,7 +22,6 @@
  * SOFTWARE.
  */
 #include "NEON/Accessor.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -30,6 +29,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/Sobel5x5.cpp b/tests/validation_old/NEON/Sobel5x5.cpp
index 2f26e62e83..e9835b6095 100644
--- a/tests/validation_old/NEON/Sobel5x5.cpp
+++ b/tests/validation_old/NEON/Sobel5x5.cpp
@@ -24,7 +24,6 @@
 
 #include "NEON/Accessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -32,6 +31,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/Threshold.cpp b/tests/validation_old/NEON/Threshold.cpp
index d56ec5eb42..6f42f19d84 100644
--- a/tests/validation_old/NEON/Threshold.cpp
+++ b/tests/validation_old/NEON/Threshold.cpp
@@ -23,7 +23,6 @@
  */
 #include "NEON/Accessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -31,6 +30,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/dataset/ThresholdDataset.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/NEON/WarpPerspective.cpp b/tests/validation_old/NEON/WarpPerspective.cpp
index 5a15591261..97bda1eed8 100644
--- a/tests/validation_old/NEON/WarpPerspective.cpp
+++ b/tests/validation_old/NEON/WarpPerspective.cpp
@@ -23,7 +23,6 @@
  */
 #include "NEON/Accessor.h"
 #include "PaddingCalculator.h"
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/Globals.h"
@@ -32,6 +31,7 @@
 #include "tests/validation_old/Reference.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/ProgramOptions.cpp b/tests/validation_old/ProgramOptions.cpp
index 08375b344c..96320e80f9 100644
--- a/tests/validation_old/ProgramOptions.cpp
+++ b/tests/validation_old/ProgramOptions.cpp
@@ -24,8 +24,8 @@
 #include "ProgramOptions.h"
 
 #include "arm_compute/core/Types.h"
-#include "tests/TypePrinter.h"
 #include "tests/TypeReader.h"
+#include "utils/TypePrinter.h"
 
 #include <random>
 #include <sstream>
diff --git a/tests/validation_old/ReferenceCPP.cpp b/tests/validation_old/ReferenceCPP.cpp
index eae892af26..cca8997485 100644
--- a/tests/validation_old/ReferenceCPP.cpp
+++ b/tests/validation_old/ReferenceCPP.cpp
@@ -26,13 +26,13 @@
 #include "TensorFactory.h"
 #include "TensorOperations.h"
 #include "TensorVisitors.h"
-#include "TypePrinter.h"
 
 #include "arm_compute/core/Coordinates.h"
 #include "arm_compute/core/Error.h"
 #include "arm_compute/core/TensorInfo.h"
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/runtime/Tensor.h"
+#include "utils/TypePrinter.h"
 
 #include "tests/validation_old/boost_wrapper.h"
 
diff --git a/tests/validation_old/UNIT/FixedPoint.cpp b/tests/validation_old/UNIT/FixedPoint.cpp
index 53fef97769..6a92cfb963 100644
--- a/tests/validation_old/UNIT/FixedPoint.cpp
+++ b/tests/validation_old/UNIT/FixedPoint.cpp
@@ -23,11 +23,11 @@
  */
 #include "tests/validation_old/FixedPoint.h"
 
-#include "TypePrinter.h"
 #include "Utils.h"
 #include "support/ToolchainSupport.h"
 #include "tests/validation_old/Validation.h"
 #include "tests/validation_old/ValidationUserConfiguration.h"
+#include "utils/TypePrinter.h"
 
 #include "tests/validation_old/boost_wrapper.h"
 
diff --git a/tests/validation_old/UNIT/TensorInfo.cpp b/tests/validation_old/UNIT/TensorInfo.cpp
index 488378ac1d..ec1db42e97 100644
--- a/tests/validation_old/UNIT/TensorInfo.cpp
+++ b/tests/validation_old/UNIT/TensorInfo.cpp
@@ -21,8 +21,8 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#include "TypePrinter.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorInfo.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/UNIT/TensorShape.cpp b/tests/validation_old/UNIT/TensorShape.cpp
index dc75b93de4..3505b1ae05 100644
--- a/tests/validation_old/UNIT/TensorShape.cpp
+++ b/tests/validation_old/UNIT/TensorShape.cpp
@@ -21,8 +21,8 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#include "TypePrinter.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 
diff --git a/tests/validation_old/UNIT/Utils.cpp b/tests/validation_old/UNIT/Utils.cpp
index b302bf27c1..a32fe189ee 100644
--- a/tests/validation_old/UNIT/Utils.cpp
+++ b/tests/validation_old/UNIT/Utils.cpp
@@ -23,8 +23,8 @@
  */
 #include "Utils.h"
 
-#include "TypePrinter.h"
 #include "tests/validation_old/Validation.h"
+#include "utils/TypePrinter.h"
 
 #include "tests/validation_old/boost_wrapper.h"
 
diff --git a/tests/validation_old/Validation.cpp b/tests/validation_old/Validation.cpp
index 8f173ba962..da5e874b24 100644
--- a/tests/validation_old/Validation.cpp
+++ b/tests/validation_old/Validation.cpp
@@ -31,9 +31,9 @@
 #include "arm_compute/runtime/Tensor.h"
 #include "tests/IAccessor.h"
 #include "tests/RawTensor.h"
-#include "tests/TypePrinter.h"
 #include "tests/Utils.h"
 #include "tests/validation_old/half.h"
+#include "utils/TypePrinter.h"
 
 #include <array>
 #include <cmath>
diff --git a/tests/validation_old/dataset/ActivationLayerDataset.h b/tests/validation_old/dataset/ActivationLayerDataset.h
index ead52a2961..881e19ff79 100644
--- a/tests/validation_old/dataset/ActivationLayerDataset.h
+++ b/tests/validation_old/dataset/ActivationLayerDataset.h
@@ -24,7 +24,7 @@
 #ifndef __ARM_COMPUTE_TEST_DATASET_ACTIVATION_LAYER_DATASET_H__
 #define __ARM_COMPUTE_TEST_DATASET_ACTIVATION_LAYER_DATASET_H__
 
-#include "TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/dataset/BatchNormalizationLayerDataset.h b/tests/validation_old/dataset/BatchNormalizationLayerDataset.h
index ca1e3b694c..3ad1befe33 100644
--- a/tests/validation_old/dataset/BatchNormalizationLayerDataset.h
+++ b/tests/validation_old/dataset/BatchNormalizationLayerDataset.h
@@ -24,7 +24,7 @@
 #ifndef __ARM_COMPUTE_TEST_DATASET_BATCH_NORMALIZATION_LAYER_DATASET_H__
 #define __ARM_COMPUTE_TEST_DATASET_BATCH_NORMALIZATION_LAYER_DATASET_H__
 
-#include "TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/dataset/ConvolutionLayerDataset.h b/tests/validation_old/dataset/ConvolutionLayerDataset.h
index 4fcba8d86d..e612c3ec93 100644
--- a/tests/validation_old/dataset/ConvolutionLayerDataset.h
+++ b/tests/validation_old/dataset/ConvolutionLayerDataset.h
@@ -24,7 +24,7 @@
 #ifndef __ARM_COMPUTE_TEST_DATASET_CONVOLUTION_LAYER_DATASET_H__
 #define __ARM_COMPUTE_TEST_DATASET_CONVOLUTION_LAYER_DATASET_H__
 
-#include "TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "tests/validation_old/dataset/GenericDataset.h"
diff --git a/tests/validation_old/dataset/FullyConnectedLayerDataset.h b/tests/validation_old/dataset/FullyConnectedLayerDataset.h
index 3564560788..8a8fed8f0f 100644
--- a/tests/validation_old/dataset/FullyConnectedLayerDataset.h
+++ b/tests/validation_old/dataset/FullyConnectedLayerDataset.h
@@ -24,7 +24,7 @@
 #ifndef __ARM_COMPUTE_TEST_DATASET_FULLY_CONNECTED_LAYER_DATASET_H__
 #define __ARM_COMPUTE_TEST_DATASET_FULLY_CONNECTED_LAYER_DATASET_H__
 
-#include "TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "tests/validation_old/dataset/GenericDataset.h"
diff --git a/tests/validation_old/dataset/GEMMDataset.h b/tests/validation_old/dataset/GEMMDataset.h
index 5250827340..ee2a387452 100644
--- a/tests/validation_old/dataset/GEMMDataset.h
+++ b/tests/validation_old/dataset/GEMMDataset.h
@@ -24,7 +24,7 @@
 #ifndef __ARM_COMPUTE_TEST_DATASET_GEMM_DATASET_H__
 #define __ARM_COMPUTE_TEST_DATASET_GEMM_DATASET_H__
 
-#include "TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/dataset/NormalizationLayerDataset.h b/tests/validation_old/dataset/NormalizationLayerDataset.h
index cd3c14d948..489d1e0af2 100644
--- a/tests/validation_old/dataset/NormalizationLayerDataset.h
+++ b/tests/validation_old/dataset/NormalizationLayerDataset.h
@@ -24,7 +24,7 @@
 #ifndef __ARM_COMPUTE_TEST_DATASET_NORMALIZATION_LAYER_DATASET_H__
 #define __ARM_COMPUTE_TEST_DATASET_NORMALIZATION_LAYER_DATASET_H__
 
-#include "TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/dataset/PoolingLayerDataset.h b/tests/validation_old/dataset/PoolingLayerDataset.h
index 6895ae4fae..a4e570c35e 100644
--- a/tests/validation_old/dataset/PoolingLayerDataset.h
+++ b/tests/validation_old/dataset/PoolingLayerDataset.h
@@ -24,7 +24,7 @@
 #ifndef __ARM_COMPUTE_TEST_DATASET_POOLING_LAYER_DATASET_H__
 #define __ARM_COMPUTE_TEST_DATASET_POOLING_LAYER_DATASET_H__
 
-#include "TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/tests/validation_old/dataset/ThresholdDataset.h b/tests/validation_old/dataset/ThresholdDataset.h
index 74d0b9cfbe..789ffac14a 100644
--- a/tests/validation_old/dataset/ThresholdDataset.h
+++ b/tests/validation_old/dataset/ThresholdDataset.h
@@ -24,7 +24,7 @@
 #ifndef __ARM_COMPUTE_TEST_DATASET_THRESHOLD_DATASET_H__
 #define __ARM_COMPUTE_TEST_DATASET_THRESHOLD_DATASET_H__
 
-#include "TypePrinter.h"
+#include "utils/TypePrinter.h"
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
diff --git a/utils/GraphUtils.cpp b/utils/GraphUtils.cpp
new file mode 100644
index 0000000000..d763606867
--- /dev/null
+++ b/utils/GraphUtils.cpp
@@ -0,0 +1,165 @@
+/*
+ * Copyright (c) 2017 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 "utils/GraphUtils.h"
+#include "utils/Utils.h"
+
+#ifdef ARM_COMPUTE_CL
+#include "arm_compute/core/CL/OpenCL.h"
+#include "arm_compute/runtime/CL/CLTensor.h"
+#endif /* ARM_COMPUTE_CL */
+
+#include "arm_compute/core/Error.h"
+#include "libnpy/npy.hpp"
+
+#include <sstream>
+
+using namespace arm_compute::graph_utils;
+
+PPMWriter::PPMWriter(std::string name, unsigned int maximum)
+    : _name(std::move(name)), _iterator(0), _maximum(maximum)
+{
+}
+
+bool PPMWriter::access_tensor(ITensor &tensor)
+{
+    std::stringstream ss;
+    ss << _name << _iterator << ".ppm";
+    if(dynamic_cast<Tensor *>(&tensor) != nullptr)
+    {
+        arm_compute::utils::save_to_ppm(dynamic_cast<Tensor &>(tensor), ss.str());
+    }
+#ifdef ARM_COMPUTE_CL
+    else if(dynamic_cast<CLTensor *>(&tensor) != nullptr)
+    {
+        arm_compute::utils::save_to_ppm(dynamic_cast<CLTensor &>(tensor), ss.str());
+    }
+#endif /* ARM_COMPUTE_CL */
+
+    _iterator++;
+    if(_maximum == 0)
+    {
+        return true;
+    }
+    return _iterator < _maximum;
+}
+
+DummyAccessor::DummyAccessor(unsigned int maximum)
+    : _iterator(0), _maximum(maximum)
+{
+}
+
+bool DummyAccessor::access_tensor(ITensor &tensor)
+{
+    ARM_COMPUTE_UNUSED(tensor);
+    bool ret = _maximum == 0 || _iterator < _maximum;
+    if(_iterator == _maximum)
+    {
+        _iterator = 0;
+    }
+    else
+    {
+        _iterator++;
+    }
+    return ret;
+}
+
+NumPyBinLoader::NumPyBinLoader(std::string filename)
+    : _filename(std::move(filename))
+{
+}
+
+bool NumPyBinLoader::access_tensor(ITensor &tensor)
+{
+    const TensorShape          tensor_shape = tensor.info()->tensor_shape();
+    std::vector<unsigned long> shape;
+
+    // Open file
+    std::ifstream stream(_filename, std::ios::in | std::ios::binary);
+    ARM_COMPUTE_ERROR_ON_MSG(!stream.good(), "Failed to load binary data");
+    // Check magic bytes and version number
+    unsigned char v_major = 0;
+    unsigned char v_minor = 0;
+    npy::read_magic(stream, &v_major, &v_minor);
+
+    // Read header
+    std::string header;
+    if(v_major == 1 && v_minor == 0)
+    {
+        header = npy::read_header_1_0(stream);
+    }
+    else if(v_major == 2 && v_minor == 0)
+    {
+        header = npy::read_header_2_0(stream);
+    }
+    else
+    {
+        ARM_COMPUTE_ERROR("Unsupported file format version");
+    }
+
+    // Parse header
+    bool        fortran_order = false;
+    std::string typestr;
+    npy::ParseHeader(header, typestr, &fortran_order, shape);
+
+    // Check if the typestring matches the given one
+    std::string expect_typestr = arm_compute::utils::get_typestring(tensor.info()->data_type());
+    ARM_COMPUTE_ERROR_ON_MSG(typestr != expect_typestr, "Typestrings mismatch");
+
+    // Validate tensor shape
+    ARM_COMPUTE_ERROR_ON_MSG(shape.size() != tensor_shape.num_dimensions(), "Tensor ranks mismatch");
+    if(fortran_order)
+    {
+        for(size_t i = 0; i < shape.size(); ++i)
+        {
+            ARM_COMPUTE_ERROR_ON_MSG(tensor_shape[i] != shape[i], "Tensor dimensions mismatch");
+        }
+    }
+    else
+    {
+        for(size_t i = 0; i < shape.size(); ++i)
+        {
+            ARM_COMPUTE_ERROR_ON_MSG(tensor_shape[i] != shape[shape.size() - i - 1], "Tensor dimensions mismatch");
+        }
+    }
+
+    // Read data
+    if(tensor.info()->padding().empty())
+    {
+        // If tensor has no padding read directly from stream.
+        stream.read(reinterpret_cast<char *>(tensor.buffer()), tensor.info()->total_size());
+    }
+    else
+    {
+        // If tensor has padding accessing tensor elements through execution window.
+        Window window;
+        window.use_tensor_dimensions(tensor_shape);
+
+        execute_window_loop(window, [&](const Coordinates & id)
+        {
+            stream.read(reinterpret_cast<char *>(tensor.ptr_to_element(id)), tensor.info()->element_size());
+        });
+    }
+    return true;
+}
diff --git a/utils/GraphUtils.h b/utils/GraphUtils.h
new file mode 100644
index 0000000000..a19f7e510d
--- /dev/null
+++ b/utils/GraphUtils.h
@@ -0,0 +1,97 @@
+/*
+ * Copyright (c) 2017 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_GRAPH_UTILS_H__
+#define __ARM_COMPUTE_GRAPH_UTILS_H__
+
+#include "arm_compute/graph/ITensorAccessor.h"
+#include "arm_compute/graph/Types.h"
+
+namespace arm_compute
+{
+namespace graph_utils
+{
+/** PPM writer class */
+class PPMWriter : public graph::ITensorAccessor
+{
+public:
+    /** Constructor
+     *
+     * @param[in] name    PPM file name
+     * @param[in] maximum Maximum elements to access
+     */
+    PPMWriter(std::string name, unsigned int maximum = 1);
+    /** Allows instances to move constructed */
+    PPMWriter(PPMWriter &&) = default;
+
+    // Inherited methods overriden:
+    bool access_tensor(ITensor &tensor) override;
+
+private:
+    const std::string _name;
+    unsigned int      _iterator;
+    unsigned int      _maximum;
+};
+
+/** Dummy accessor class */
+class DummyAccessor : public graph::ITensorAccessor
+{
+public:
+    /** Constructor
+     *
+     * @param[in] maximum Maximum elements to write
+     */
+    DummyAccessor(unsigned int maximum = 1);
+    /** Allows instances to move constructed */
+    DummyAccessor(DummyAccessor &&) = default;
+
+    // Inherited methods overriden:
+    bool access_tensor(ITensor &tensor) override;
+
+private:
+    unsigned int _iterator;
+    unsigned int _maximum;
+};
+
+/** Numpy Binary loader class*/
+class NumPyBinLoader : public graph::ITensorAccessor
+{
+public:
+    /** Default Constructor
+     *
+     * @param filename Binary file name
+     */
+    NumPyBinLoader(std::string filename);
+    /** Allows instances to move constructed */
+    NumPyBinLoader(NumPyBinLoader &&) = default;
+
+    // Inherited methods overriden:
+    bool access_tensor(ITensor &tensor) override;
+
+private:
+    const std::string _filename;
+};
+} // namespace graph
+} // namespace arm_compute
+
+#endif /* __ARM_COMPUTE_GRAPH_UTILS_H__ */
diff --git a/tests/TypePrinter.h b/utils/TypePrinter.h
index d3d9f8f5ac..30592fff55 100644
--- a/tests/TypePrinter.h
+++ b/utils/TypePrinter.h
@@ -51,83 +51,6 @@ inline ::std::ostream &operator<<(::std::ostream &os, const Dimensions<T> &dimen
     return os;
 }
 
-//FIXME: Check why this doesn't work and the TensorShape overload is needed
-template <typename T>
-inline std::string to_string(const Dimensions<T> &dimensions)
-{
-    std::stringstream str;
-    str << dimensions;
-    return str.str();
-}
-
-inline std::string to_string(const TensorShape &shape)
-{
-    std::stringstream str;
-    str << shape;
-    return str.str();
-}
-
-/** Formatted output of the Rectangle type. */
-inline ::std::ostream &operator<<(::std::ostream &os, const Rectangle &rect)
-{
-    os << rect.width << "x" << rect.height;
-    os << "+" << rect.x << "+" << rect.y;
-
-    return os;
-}
-
-/** Formatted output of the PadStridInfo type. */
-inline ::std::ostream &operator<<(::std::ostream &os, const PadStrideInfo &pad_stride_info)
-{
-    os << pad_stride_info.stride().first << "," << pad_stride_info.stride().second;
-    os << ";";
-    os << pad_stride_info.pad().first << "," << pad_stride_info.pad().second;
-
-    return os;
-}
-
-inline std::string to_string(const PadStrideInfo &pad_stride_info)
-{
-    std::stringstream str;
-    str << pad_stride_info;
-    return str.str();
-}
-
-/** Formatted output of the ROIPoolingInfo type. */
-inline ::std::ostream &operator<<(::std::ostream &os, const ROIPoolingLayerInfo &pool_info)
-{
-    os << pool_info.pooled_width() << "x" << pool_info.pooled_height() << "~" << pool_info.spatial_scale();
-    return os;
-}
-
-/** Formatted output of the BorderMode type. */
-inline ::std::ostream &operator<<(::std::ostream &os, const BorderMode &mode)
-{
-    switch(mode)
-    {
-        case BorderMode::UNDEFINED:
-            os << "UNDEFINED";
-            break;
-        case BorderMode::CONSTANT:
-            os << "CONSTANT";
-            break;
-        case BorderMode::REPLICATE:
-            os << "REPLICATE";
-            break;
-        default:
-            ARM_COMPUTE_ERROR("NOT_SUPPORTED!");
-    }
-
-    return os;
-}
-
-inline std::string to_string(const BorderMode &mode)
-{
-    std::stringstream str;
-    str << mode;
-    return str.str();
-}
-
 /** Formatted output of the NonLinearFilterFunction type. */
 inline ::std::ostream &operator<<(::std::ostream &os, const NonLinearFilterFunction &function)
 {
@@ -187,19 +110,19 @@ inline std::string to_string(const MatrixPattern &pattern)
     return str.str();
 }
 
-/** Formatted output of the InterpolationPolicy type. */
-inline ::std::ostream &operator<<(::std::ostream &os, const InterpolationPolicy &policy)
+/** Formatted output of the RoundingPolicy type. */
+inline ::std::ostream &operator<<(::std::ostream &os, const RoundingPolicy &rounding_policy)
 {
-    switch(policy)
+    switch(rounding_policy)
     {
-        case InterpolationPolicy::NEAREST_NEIGHBOR:
-            os << "NEAREST_NEIGHBOR";
+        case RoundingPolicy::TO_ZERO:
+            os << "TO_ZERO";
             break;
-        case InterpolationPolicy::BILINEAR:
-            os << "BILINEAR";
+        case RoundingPolicy::TO_NEAREST_UP:
+            os << "TO_NEAREST_UP";
             break;
-        case InterpolationPolicy::AREA:
-            os << "AREA";
+        case RoundingPolicy::TO_NEAREST_EVEN:
+            os << "TO_NEAREST_EVEN";
             break;
         default:
             ARM_COMPUTE_ERROR("NOT_SUPPORTED!");
@@ -208,60 +131,22 @@ inline ::std::ostream &operator<<(::std::ostream &os, const InterpolationPolicy
     return os;
 }
 
-inline std::string to_string(const InterpolationPolicy &policy)
+/** Formatted output of the WeightsInfo type. */
+inline ::std::ostream &operator<<(::std::ostream &os, const WeightsInfo &weights_info)
 {
-    std::stringstream str;
-    str << policy;
-    return str.str();
-}
-
-/** Formatted output of the ConversionPolicy type. */
-inline ::std::ostream &operator<<(::std::ostream &os, const ConvertPolicy &policy)
-{
-    switch(policy)
-    {
-        case ConvertPolicy::WRAP:
-            os << "WRAP";
-            break;
-        case ConvertPolicy::SATURATE:
-            os << "SATURATE";
-            break;
-        default:
-            ARM_COMPUTE_ERROR("NOT_SUPPORTED!");
-    }
+    os << weights_info.are_reshaped() << ";";
+    os << weights_info.num_kernels() << ";" << weights_info.kernel_size().first << "," << weights_info.kernel_size().second;
 
     return os;
 }
 
-inline std::string to_string(const ConvertPolicy &policy)
-{
-    std::stringstream str;
-    str << policy;
-    return str.str();
-}
-
-/** Formatted output of the Reduction Operations. */
-inline ::std::ostream &operator<<(::std::ostream &os, const ReductionOperation &op)
+/** Formatted output of the ROIPoolingInfo type. */
+inline ::std::ostream &operator<<(::std::ostream &os, const ROIPoolingLayerInfo &pool_info)
 {
-    switch(op)
-    {
-        case ReductionOperation::SUM_SQUARE:
-            os << "SUM_SQUARE";
-            break;
-        default:
-            ARM_COMPUTE_ERROR("NOT_SUPPORTED!");
-    }
-
+    os << pool_info.pooled_width() << "x" << pool_info.pooled_height() << "~" << pool_info.spatial_scale();
     return os;
 }
 
-inline std::string to_string(const ReductionOperation &op)
-{
-    std::stringstream str;
-    str << op;
-    return str.str();
-}
-
 /** Formatted output of the activation function type. */
 inline ::std::ostream &operator<<(::std::ostream &os, const ActivationLayerInfo::ActivationFunction &act_function)
 {
@@ -282,9 +167,6 @@ inline ::std::ostream &operator<<(::std::ostream &os, const ActivationLayerInfo:
         case ActivationLayerInfo::ActivationFunction::BOUNDED_RELU:
             os << "BOUNDED_RELU";
             break;
-        case ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU:
-            os << "LU_BOUNDED_RELU";
-            break;
         case ActivationLayerInfo::ActivationFunction::LEAKY_RELU:
             os << "LEAKY_RELU";
             break;
@@ -294,6 +176,8 @@ inline ::std::ostream &operator<<(::std::ostream &os, const ActivationLayerInfo:
         case ActivationLayerInfo::ActivationFunction::SQRT:
             os << "SQRT";
             break;
+        case ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU:
+            os << "LU_BOUNDED_RELU";
         case ActivationLayerInfo::ActivationFunction::SQUARE:
             os << "SQUARE";
             break;
@@ -307,17 +191,17 @@ inline ::std::ostream &operator<<(::std::ostream &os, const ActivationLayerInfo:
     return os;
 }
 
-inline std::string to_string(const ActivationLayerInfo::ActivationFunction &function)
+inline std::string to_string(const arm_compute::ActivationLayerInfo &info)
 {
     std::stringstream str;
-    str << function;
+    str << info.activation();
     return str.str();
 }
 
-inline std::string to_string(const ActivationLayerInfo &info)
+inline std::string to_string(const arm_compute::ActivationLayerInfo::ActivationFunction &function)
 {
     std::stringstream str;
-    str << info.activation();
+    str << function;
     return str.str();
 }
 
@@ -342,14 +226,7 @@ inline ::std::ostream &operator<<(::std::ostream &os, const NormType &norm_type)
     return os;
 }
 
-inline std::string to_string(const NormType &type)
-{
-    std::stringstream str;
-    str << type;
-    return str.str();
-}
-
-inline std::string to_string(const NormalizationLayerInfo &info)
+inline std::string to_string(const arm_compute::NormalizationLayerInfo &info)
 {
     std::stringstream str;
     str << info.type();
@@ -377,45 +254,10 @@ inline ::std::ostream &operator<<(::std::ostream &os, const PoolingType &pool_ty
     return os;
 }
 
-inline std::string to_string(const PoolingType &type)
-{
-    std::stringstream str;
-    str << type;
-    return str.str();
-}
-
 /** Formatted output of @ref PoolingLayerInfo. */
 inline ::std::ostream &operator<<(::std::ostream &os, const PoolingLayerInfo &info)
 {
-    os << info.pool_type() << ";" << info.pool_size() << ";" << info.pad_stride_info();
-
-    return os;
-}
-
-inline std::string to_string(const PoolingLayerInfo &info)
-{
-    std::stringstream str;
-    str << info.pool_type();
-    return str.str();
-}
-
-/** Formatted output of the RoundingPolicy type. */
-inline ::std::ostream &operator<<(::std::ostream &os, const RoundingPolicy &rounding_policy)
-{
-    switch(rounding_policy)
-    {
-        case RoundingPolicy::TO_ZERO:
-            os << "TO_ZERO";
-            break;
-        case RoundingPolicy::TO_NEAREST_UP:
-            os << "TO_NEAREST_UP";
-            break;
-        case RoundingPolicy::TO_NEAREST_EVEN:
-            os << "TO_NEAREST_EVEN";
-            break;
-        default:
-            ARM_COMPUTE_ERROR("NOT_SUPPORTED!");
-    }
+    os << info.pool_type();
 
     return os;
 }
@@ -477,7 +319,7 @@ inline ::std::ostream &operator<<(::std::ostream &os, const DataType &data_type)
     return os;
 }
 
-inline std::string to_string(const DataType &data_type)
+inline std::string to_string(const arm_compute::DataType &data_type)
 {
     std::stringstream str;
     str << data_type;
@@ -595,6 +437,27 @@ inline ::std::ostream &operator<<(::std::ostream &os, const Channel &channel)
     return os;
 }
 
+/** Formatted output of the BorderMode type. */
+inline ::std::ostream &operator<<(::std::ostream &os, const BorderMode &mode)
+{
+    switch(mode)
+    {
+        case BorderMode::UNDEFINED:
+            os << "UNDEFINED";
+            break;
+        case BorderMode::CONSTANT:
+            os << "CONSTANT";
+            break;
+        case BorderMode::REPLICATE:
+            os << "REPLICATE";
+            break;
+        default:
+            ARM_COMPUTE_ERROR("NOT_SUPPORTED!");
+    }
+
+    return os;
+}
+
 /** Formatted output of the BorderSize type. */
 inline ::std::ostream &operator<<(::std::ostream &os, const BorderSize &border)
 {
@@ -605,5 +468,152 @@ inline ::std::ostream &operator<<(::std::ostream &os, const BorderSize &border)
 
     return os;
 }
+
+/** Formatted output of the InterpolationPolicy type. */
+inline ::std::ostream &operator<<(::std::ostream &os, const InterpolationPolicy &policy)
+{
+    switch(policy)
+    {
+        case InterpolationPolicy::NEAREST_NEIGHBOR:
+            os << "NEAREST_NEIGHBOR";
+            break;
+        case InterpolationPolicy::BILINEAR:
+            os << "BILINEAR";
+            break;
+        case InterpolationPolicy::AREA:
+            os << "AREA";
+            break;
+        default:
+            ARM_COMPUTE_ERROR("NOT_SUPPORTED!");
+    }
+
+    return os;
+}
+
+//FIXME: Check why this doesn't work and the TensorShape overload is needed
+template <typename T>
+inline std::string to_string(const Dimensions<T> &dimensions)
+{
+    std::stringstream str;
+    str << dimensions;
+    return str.str();
+}
+
+/** Formatted output of the TensorShape type. */
+inline std::string to_string(const TensorShape &shape)
+{
+    std::stringstream str;
+    str << shape;
+    return str.str();
+}
+
+/** Formatted output of the Rectangle type. */
+inline ::std::ostream &operator<<(::std::ostream &os, const Rectangle &rect)
+{
+    os << rect.width << "x" << rect.height;
+    os << "+" << rect.x << "+" << rect.y;
+
+    return os;
+}
+
+/** Formatted output of the PadStridInfo type. */
+inline ::std::ostream &operator<<(::std::ostream &os, const PadStrideInfo &pad_stride_info)
+{
+    os << pad_stride_info.stride().first << "," << pad_stride_info.stride().second;
+    os << ";";
+    os << pad_stride_info.pad().first << "," << pad_stride_info.pad().second;
+
+    return os;
+}
+
+inline std::string to_string(const PadStrideInfo &pad_stride_info)
+{
+    std::stringstream str;
+    str << pad_stride_info;
+    return str.str();
+}
+
+inline std::string to_string(const BorderMode &mode)
+{
+    std::stringstream str;
+    str << mode;
+    return str.str();
+}
+
+inline std::string to_string(const InterpolationPolicy &policy)
+{
+    std::stringstream str;
+    str << policy;
+    return str.str();
+}
+
+/** Formatted output of the ConversionPolicy type. */
+inline ::std::ostream &operator<<(::std::ostream &os, const ConvertPolicy &policy)
+{
+    switch(policy)
+    {
+        case ConvertPolicy::WRAP:
+            os << "WRAP";
+            break;
+        case ConvertPolicy::SATURATE:
+            os << "SATURATE";
+            break;
+        default:
+            ARM_COMPUTE_ERROR("NOT_SUPPORTED!");
+    }
+
+    return os;
+}
+
+inline std::string to_string(const ConvertPolicy &policy)
+{
+    std::stringstream str;
+    str << policy;
+    return str.str();
+}
+
+/** Formatted output of the Reduction Operations. */
+inline ::std::ostream &operator<<(::std::ostream &os, const ReductionOperation &op)
+{
+    switch(op)
+    {
+        case ReductionOperation::SUM_SQUARE:
+            os << "SUM_SQUARE";
+            break;
+        default:
+            ARM_COMPUTE_ERROR("NOT_SUPPORTED!");
+    }
+
+    return os;
+}
+
+inline std::string to_string(const ReductionOperation &op)
+{
+    std::stringstream str;
+    str << op;
+    return str.str();
+}
+
+inline std::string to_string(const NormType &type)
+{
+    std::stringstream str;
+    str << type;
+    return str.str();
+}
+
+inline std::string to_string(const PoolingType &type)
+{
+    std::stringstream str;
+    str << type;
+    return str.str();
+}
+
+inline std::string to_string(const PoolingLayerInfo &info)
+{
+    std::stringstream str;
+    str << info.pool_type();
+    return str.str();
+}
+
 } // namespace arm_compute
 #endif /* __ARM_COMPUTE_TEST_TYPE_PRINTER_H__ */
diff --git a/utils/Utils.h b/utils/Utils.h
index d535124598..47a352e905 100644
--- a/utils/Utils.h
+++ b/utils/Utils.h
@@ -30,6 +30,7 @@
 #include "arm_compute/core/Validate.h"
 #include "arm_compute/core/Window.h"
 #include "arm_compute/runtime/Tensor.h"
+#include "support/ToolchainSupport.h"
 
 #ifdef ARM_COMPUTE_CL
 #include "arm_compute/core/CL/OpenCL.h"
@@ -401,6 +402,56 @@ void load_trained_data(T &tensor, const std::string &filename)
     }
 }
 
+/** Obtain numpy type string from DataType.
+ *
+ * @param[in] data_type Data type.
+ *
+ * @return numpy type string.
+ */
+inline std::string get_typestring(DataType data_type)
+{
+    // Check endianness
+    const unsigned int i = 1;
+    const char        *c = reinterpret_cast<const char *>(&i);
+    std::string        endianness;
+    if(*c == 1)
+    {
+        endianness = std::string("<");
+    }
+    else
+    {
+        endianness = std::string(">");
+    }
+    const std::string no_endianness("|");
+
+    switch(data_type)
+    {
+        case DataType::U8:
+            return no_endianness + "u" + support::cpp11::to_string(sizeof(uint8_t));
+        case DataType::S8:
+            return no_endianness + "i" + support::cpp11::to_string(sizeof(int8_t));
+        case DataType::U16:
+            return endianness + "u" + support::cpp11::to_string(sizeof(uint16_t));
+        case DataType::S16:
+            return endianness + "i" + support::cpp11::to_string(sizeof(int16_t));
+        case DataType::U32:
+            return endianness + "u" + support::cpp11::to_string(sizeof(uint32_t));
+        case DataType::S32:
+            return endianness + "i" + support::cpp11::to_string(sizeof(int32_t));
+        case DataType::U64:
+            return endianness + "u" + support::cpp11::to_string(sizeof(uint64_t));
+        case DataType::S64:
+            return endianness + "i" + support::cpp11::to_string(sizeof(int64_t));
+        case DataType::F32:
+            return endianness + "f" + support::cpp11::to_string(sizeof(float));
+        case DataType::F64:
+            return endianness + "f" + support::cpp11::to_string(sizeof(double));
+        case DataType::SIZET:
+            return endianness + "u" + support::cpp11::to_string(sizeof(size_t));
+        default:
+            ARM_COMPUTE_ERROR("NOT SUPPORTED!");
+    }
+}
 } // namespace utils
 } // namespace arm_compute
 #endif /* __UTILS_UTILS_H__*/