COMPMID-897 Merge batch normalization with bounded relu

Change-Id: I9a607fe620f795cdea1a99fdd3f5f8c2fc76f980
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/119234
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>

7 files changed, 125 insertions, 100 deletions

diff --git a/arm_compute/core/CL/kernels/CLBatchNormalizationLayerKernel.h b/arm_compute/core/CL/kernels/CLBatchNormalizationLayerKernel.h
index 8643d83bcc..fee5dd3bae 100644
--- a/arm_compute/core/CL/kernels/CLBatchNormalizationLayerKernel.h
+++ b/arm_compute/core/CL/kernels/CLBatchNormalizationLayerKernel.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -52,35 +52,38 @@ public:
      *
      * @note If the output tensor is a nullptr, the batch normalization function will be performed in-place
      *
-     * @param[in, out] input   Source tensor. In case of @p output tensor = nullptr, this tensor will store the result.
-     *                         3 lower dimensions represent a single input with dimensions [width, height, FM].
-     *                         The rest are optional and used for representing batches. Data types supported: QS8/QS16/F16/F32.
-     * @param[out]     output  Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input
-     * @param[in]      mean    Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]      var     Variance values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]      beta    Beta values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]      gamma   Gamma values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]      epsilon Small value to avoid division with zero.
+     * @param[in, out] input    Source tensor. In case of @p output tensor = nullptr, this tensor will store the result.
+     *                          3 lower dimensions represent a single input with dimensions [width, height, FM].
+     *                          The rest are optional and used for representing batches. Data types supported: QS8/QS16/F16/F32.
+     * @param[out]     output   Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input
+     * @param[in]      mean     Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]      var      Variance values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]      beta     Beta values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]      gamma    Gamma values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]      epsilon  Small value to avoid division with zero.
+     * @param[in]      act_info (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      */
-    void configure(ICLTensor *input, ICLTensor *output, const ICLTensor *mean, const ICLTensor *var, const ICLTensor *beta, const ICLTensor *gamma, float epsilon);
+    void configure(ICLTensor *input, ICLTensor *output, const ICLTensor *mean, const ICLTensor *var, const ICLTensor *beta, const ICLTensor *gamma, float epsilon,
+                   ActivationLayerInfo act_info = ActivationLayerInfo());
     /** Static function to check if given info will lead to a valid configuration of @ref CLBatchNormalizationLayerKernel
      *
-     * @param[in] input   Source tensor info. In case of @p output tensor info = nullptr, this tensor will store the result.
-     *                    3 lower dimensions represent a single input with dimensions [width, height, FM].
-     *                    The rest are optional and used for representing batches. Data types supported: QS8/QS16/F16/F32.
-     * @param[in] output  Destination tensor info. Output will have the same number of dimensions as input. Data type supported: same as @p input
-     * @param[in] mean    Mean values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in] var     Variance values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in] beta    Beta values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in] gamma   Gamma values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in] epsilon Small value to avoid division with zero.
+     * @param[in] input    Source tensor info. In case of @p output tensor info = nullptr, this tensor will store the result.
+     *                     3 lower dimensions represent a single input with dimensions [width, height, FM].
+     *                     The rest are optional and used for representing batches. Data types supported: QS8/QS16/F16/F32.
+     * @param[in] output   Destination tensor info. Output will have the same number of dimensions as input. Data type supported: same as @p input
+     * @param[in] mean     Mean values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in] var      Variance values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in] beta     Beta values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in] gamma    Gamma values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in] epsilon  Small value to avoid division with zero.
+     * @param[in] act_info (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      *
      * @return a status
      */
     static Status validate(const ITensorInfo *input, const ITensorInfo *output,
                            const ITensorInfo *mean, const ITensorInfo *var,
                            const ITensorInfo *beta, const ITensorInfo *gamma,
-                           float epsilon);
+                           float epsilon, ActivationLayerInfo act_info);
 
     // Inherited methods overridden:
     void run(const Window &window, cl::CommandQueue &queue) override;
diff --git a/arm_compute/core/GLES_COMPUTE/kernels/GCBatchNormalizationLayerKernel.h b/arm_compute/core/GLES_COMPUTE/kernels/GCBatchNormalizationLayerKernel.h
index 2bbd6a83fe..15d7f79afb 100644
--- a/arm_compute/core/GLES_COMPUTE/kernels/GCBatchNormalizationLayerKernel.h
+++ b/arm_compute/core/GLES_COMPUTE/kernels/GCBatchNormalizationLayerKernel.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -50,16 +50,18 @@ public:
 
     /** Set the input and output tensors.
      *
-     * @param[in]  input   Source tensor. 3 lower dimensions represent a single input with dimensions [width, height, FM].
-     *                     The rest are optional and used for representing batches. Data types supported: F16/F32.
-     * @param[out] output  Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input
-     * @param[in]  mean    Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]  var     Variance values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]  beta    Beta values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]  gamma   Gamma values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]  epsilon Small value to avoid division with zero.
+     * @param[in]  input    Source tensor. 3 lower dimensions represent a single input with dimensions [width, height, FM].
+     *                      The rest are optional and used for representing batches. Data types supported: F16/F32.
+     * @param[out] output   Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input
+     * @param[in]  mean     Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]  var      Variance values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]  beta     Beta values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]  gamma    Gamma values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]  epsilon  Small value to avoid division with zero.
+     * @param[in]  act_info (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      */
-    void configure(const IGCTensor *input, IGCTensor *output, const IGCTensor *mean, const IGCTensor *var, const IGCTensor *beta, const IGCTensor *gamma, float epsilon);
+    void configure(const IGCTensor *input, IGCTensor *output, const IGCTensor *mean, const IGCTensor *var, const IGCTensor *beta, const IGCTensor *gamma, float epsilon,
+                   ActivationLayerInfo act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
     void run(const Window &window) override;
diff --git a/arm_compute/core/Types.h b/arm_compute/core/Types.h
index 5a08ac9153..3affe7e8ec 100644
--- a/arm_compute/core/Types.h
+++ b/arm_compute/core/Types.h
@@ -713,6 +713,7 @@ public:
         LINEAR           /**< Linear ( \f$ f(x)= ax + b \f$ ) */
     };
 
+    ActivationLayerInfo() = default;
     /** Default Constructor
      *
      * @param[in] f The activation function to use.
@@ -721,7 +722,7 @@ public:
      * @param[in] b (Optional) The beta parameter used by some activation functions (@ref ActivationFunction::LINEAR, @ref ActivationFunction::LU_BOUNDED_RELU, @ref ActivationFunction::TANH).
      */
     ActivationLayerInfo(ActivationFunction f, float a = 0.0f, float b = 0.0f)
-        : _act(f), _a(a), _b(b)
+        : _act(f), _a(a), _b(b), _enabled(true)
     {
     }
     ActivationFunction activation() const
@@ -736,11 +737,16 @@ public:
     {
         return _b;
     }
+    bool enabled() const
+    {
+        return _enabled;
+    }
 
 private:
-    ActivationFunction _act;
-    float              _a;
-    float              _b;
+    ActivationFunction _act     = { ActivationLayerInfo::ActivationFunction::LOGISTIC };
+    float              _a       = {};
+    float              _b       = {};
+    bool               _enabled = { false };
 };
 
 /** Normalization Layer Information class */
diff --git a/arm_compute/graph/nodes/BatchNormalizationLayer.h b/arm_compute/graph/nodes/BatchNormalizationLayer.h
index df7b1d19a9..266c3905d8 100644
--- a/arm_compute/graph/nodes/BatchNormalizationLayer.h
+++ b/arm_compute/graph/nodes/BatchNormalizationLayer.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -40,15 +40,16 @@ class BatchNormalizationLayer final : public INode
 public:
     /** Default constructor
      *
-     * @param[in] mean    Mean values tensor
-     * @param[in] var     Var values tensor
-     * @param[in] gamma   Gamma values tensor
-     * @param[in] beta    Beta values tensor
-     * @param[in] epsilon Epsilon value
+     * @param[in] mean     Mean values tensor
+     * @param[in] var      Var values tensor
+     * @param[in] gamma    Gamma values tensor
+     * @param[in] beta     Beta values tensor
+     * @param[in] epsilon  Epsilon value
+     * @param[in] act_info (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      */
     template <typename AccessorType>
-    BatchNormalizationLayer(AccessorType &&mean, AccessorType &&var, AccessorType &&gamma, AccessorType &&beta, float epsilon)
-        : _mean(std::move(mean)), _var(std::move(var)), _gamma(std::move(gamma)), _beta(std::move(beta)), _epsilon(epsilon)
+    BatchNormalizationLayer(AccessorType &&mean, AccessorType &&var, AccessorType &&gamma, AccessorType &&beta, float epsilon, ActivationLayerInfo act_info = ActivationLayerInfo())
+        : _mean(std::move(mean)), _var(std::move(var)), _gamma(std::move(gamma)), _beta(std::move(beta)), _epsilon(epsilon), _act_info(act_info)
     {
     }
 
@@ -56,11 +57,12 @@ public:
     std::unique_ptr<arm_compute::IFunction> instantiate_node(GraphContext &ctx, ITensorObject *input, ITensorObject *output) override;
 
 private:
-    Tensor _mean;
-    Tensor _var;
-    Tensor _gamma;
-    Tensor _beta;
-    float  _epsilon;
+    Tensor              _mean;
+    Tensor              _var;
+    Tensor              _gamma;
+    Tensor              _beta;
+    float               _epsilon;
+    ActivationLayerInfo _act_info;
 };
 } // namespace graph
 } // namespace arm_compute
diff --git a/arm_compute/runtime/CL/functions/CLBatchNormalizationLayer.h b/arm_compute/runtime/CL/functions/CLBatchNormalizationLayer.h
index 127de10555..3d5145a697 100644
--- a/arm_compute/runtime/CL/functions/CLBatchNormalizationLayer.h
+++ b/arm_compute/runtime/CL/functions/CLBatchNormalizationLayer.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -48,35 +48,38 @@ public:
      *
      * @note If the output tensor is a nullptr, the batch normalization function will be performed in-place
      *
-     * @param[in, out] input   Source tensor. In case of @p output tensor = nullptr, this tensor will store the result.
-     *                         3 lower dimensions represent a single input with dimensions [width, height, FM].
-     *                         The rest are optional and used for representing batches. Data types supported: QS8/QS16/F16/F32.
-     * @param[out]     output  Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input
-     * @param[in]      mean    Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]      var     Variance values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]      beta    Beta values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]      gamma   Gamma values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]      epsilon Small value to avoid division with zero.
+     * @param[in, out] input    Source tensor. In case of @p output tensor = nullptr, this tensor will store the result.
+     *                          3 lower dimensions represent a single input with dimensions [width, height, FM].
+     *                          The rest are optional and used for representing batches. Data types supported: QS8/QS16/F16/F32.
+     * @param[out]     output   Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input
+     * @param[in]      mean     Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]      var      Variance values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]      beta     Beta values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]      gamma    Gamma values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]      epsilon  Small value to avoid division with zero.
+     * @param[in]      act_info (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      */
-    void configure(ICLTensor *input, ICLTensor *output, const ICLTensor *mean, const ICLTensor *var, const ICLTensor *beta, const ICLTensor *gamma, float epsilon);
+    void configure(ICLTensor *input, ICLTensor *output, const ICLTensor *mean, const ICLTensor *var, const ICLTensor *beta, const ICLTensor *gamma, float epsilon,
+                   ActivationLayerInfo act_info = ActivationLayerInfo());
     /** Static function to check if given info will lead to a valid configuration of @ref CLBatchNormalizationLayer
      *
-     * @param[in] input   Source tensor info. In case of @p output tensor info = nullptr, this tensor will store the result.
-     *                    3 lower dimensions represent a single input with dimensions [width, height, FM].
-     *                    The rest are optional and used for representing batches. Data types supported: QS8/QS16/F16/F32.
-     * @param[in] output  Destination tensor info. Output will have the same number of dimensions as input. Data type supported: same as @p input
-     * @param[in] mean    Mean values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in] var     Variance values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in] beta    Beta values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in] gamma   Gamma values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in] epsilon Small value to avoid division with zero.
+     * @param[in] input    Source tensor info. In case of @p output tensor info = nullptr, this tensor will store the result.
+     *                     3 lower dimensions represent a single input with dimensions [width, height, FM].
+     *                     The rest are optional and used for representing batches. Data types supported: QS8/QS16/F16/F32.
+     * @param[in] output   Destination tensor info. Output will have the same number of dimensions as input. Data type supported: same as @p input
+     * @param[in] mean     Mean values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in] var      Variance values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in] beta     Beta values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in] gamma    Gamma values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in] epsilon  Small value to avoid division with zero.
+     * @param[in] act_info (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      *
      * @return a status
      */
     static Status validate(const ITensorInfo *input, const ITensorInfo *output,
                            const ITensorInfo *mean, const ITensorInfo *var,
                            const ITensorInfo *beta, const ITensorInfo *gamma,
-                           float epsilon);
+                           float epsilon, ActivationLayerInfo act_info);
 
     // Inherited methods overridden:
     void run() override;
diff --git a/arm_compute/runtime/GLES_COMPUTE/functions/GCBatchNormalizationLayer.h b/arm_compute/runtime/GLES_COMPUTE/functions/GCBatchNormalizationLayer.h
index 9d81b9a7f7..01e53d26f5 100644
--- a/arm_compute/runtime/GLES_COMPUTE/functions/GCBatchNormalizationLayer.h
+++ b/arm_compute/runtime/GLES_COMPUTE/functions/GCBatchNormalizationLayer.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -46,16 +46,18 @@ public:
     GCBatchNormalizationLayer();
     /** Set the input and output tensors.
      *
-     * @param[in]  input   Source tensor. 3 lower dimensions represent a single input with dimensions [width, height, FM].
-     *                     The rest are optional and used for representing batches. Data types supported: F16/F32.
-     * @param[out] output  Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input
-     * @param[in]  mean    Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]  var     Variance values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]  beta    Beta values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]  gamma   Gamma values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]  epsilon Small value to avoid division with zero.
+     * @param[in]  input    Source tensor. 3 lower dimensions represent a single input with dimensions [width, height, FM].
+     *                      The rest are optional and used for representing batches. Data types supported: F16/F32.
+     * @param[out] output   Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input
+     * @param[in]  mean     Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]  var      Variance values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]  beta     Beta values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]  gamma    Gamma values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]  epsilon  Small value to avoid division with zero.
+     * @param[in]  act_info (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      */
-    void configure(const IGCTensor *input, IGCTensor *output, const IGCTensor *mean, const IGCTensor *var, const IGCTensor *beta, const IGCTensor *gamma, float epsilon);
+    void configure(const IGCTensor *input, IGCTensor *output, const IGCTensor *mean, const IGCTensor *var, const IGCTensor *beta, const IGCTensor *gamma, float epsilon,
+                   ActivationLayerInfo act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
     void run() override;
diff --git a/arm_compute/runtime/NEON/functions/NEBatchNormalizationLayer.h b/arm_compute/runtime/NEON/functions/NEBatchNormalizationLayer.h
index 1933468afc..5c8200beda 100644
--- a/arm_compute/runtime/NEON/functions/NEBatchNormalizationLayer.h
+++ b/arm_compute/runtime/NEON/functions/NEBatchNormalizationLayer.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -27,6 +27,7 @@
 #include "arm_compute/core/NEON/kernels/NEBatchNormalizationLayerKernel.h"
 #include "arm_compute/core/Types.h"
 #include "arm_compute/runtime/IFunction.h"
+#include "arm_compute/runtime/NEON/functions/NEActivationLayer.h"
 
 namespace arm_compute
 {
@@ -47,41 +48,47 @@ public:
      *
      * @note If the output tensor is a nullptr, the batch normalization function will be performed in-place
      *
-     * @param[in, out] input   Source tensor. In case of @p output tensor = nullptr, this tensor will store the result.
-     *                         3 lower dimensions represent a single input with dimensions [width, height, FM].
-     *                         The rest are optional and used for representing batches. Data types supported: QS8/QS16/F16/F32.
-     * @param[out]     output  Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input
-     * @param[in]      mean    Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]      var     Variance values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]      beta    Beta values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]      gamma   Gamma values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in]      epsilon Small value to avoid division with zero.
+     * @param[in, out] input    Source tensor. In case of @p output tensor = nullptr, this tensor will store the result.
+     *                          3 lower dimensions represent a single input with dimensions [width, height, FM].
+     *                          The rest are optional and used for representing batches. Data types supported: QS8/QS16/F16/F32.
+     * @param[out]     output   Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input
+     * @param[in]      mean     Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]      var      Variance values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]      beta     Beta values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]      gamma    Gamma values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in]      epsilon  Small value to avoid division with zero.
+     * @param[in]      act_info (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      */
-    void configure(ITensor *input, ITensor *output, const ITensor *mean, const ITensor *var, const ITensor *beta, const ITensor *gamma, float epsilon);
+    void configure(ITensor *input, ITensor *output, const ITensor *mean, const ITensor *var, const ITensor *beta, const ITensor *gamma, float epsilon,
+                   ActivationLayerInfo act_info = ActivationLayerInfo());
     /** Static function to check if given info will lead to a valid configuration of @ref NEBatchNormalizationLayer
      *
-     * @param[in] input   Source tensor info. In case of @p output tensor = nullptr, this tensor will store the result.
-     *                    3 lower dimensions represent a single input with dimensions [width, height, FM].
-     *                    The rest are optional and used for representing batches. Data types supported: QS8/QS16/F16/F32.
-     * @param[in] output  Destination tensor info. Output will have the same number of dimensions as input. Data type supported: same as @p input
-     * @param[in] mean    Mean values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in] var     Variance values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in] beta    Beta values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in] gamma   Gamma values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
-     * @param[in] epsilon Small value to avoid division with zero.
+     * @param[in] input    Source tensor info. In case of @p output tensor = nullptr, this tensor will store the result.
+     *                     3 lower dimensions represent a single input with dimensions [width, height, FM].
+     *                     The rest are optional and used for representing batches. Data types supported: QS8/QS16/F16/F32.
+     * @param[in] output   Destination tensor info. Output will have the same number of dimensions as input. Data type supported: same as @p input
+     * @param[in] mean     Mean values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in] var      Variance values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in] beta     Beta values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in] gamma    Gamma values tensor info. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input
+     * @param[in] epsilon  Small value to avoid division with zero.
+     * @param[in] act_info (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      *
      * @return a status
      */
     static Status validate(const ITensorInfo *input, const ITensorInfo *output,
                            const ITensorInfo *mean, const ITensorInfo *var,
                            const ITensorInfo *beta, const ITensorInfo *gamma,
-                           float epsilon);
+                           float epsilon, ActivationLayerInfo act_info);
 
     // Inherited methods overridden:
     void run() override;
 
 private:
     NEBatchNormalizationLayerKernel _norm_kernel; /**< Batch normalization layer kernel */
+    // COMPMID-906 Use fused activation in NEON Batch normalization
+    NEActivationLayer _act_func;
+    bool              _act_info_enabled;
 };
 }
 #endif /* __ARM_COMPUTE_NEBATCHNORMALIZATIONLAYER_H__ */