arm_compute v19.05
diff --git a/arm_compute/graph/Graph.h b/arm_compute/graph/Graph.h
index 2a77682..878976f 100644
--- a/arm_compute/graph/Graph.h
+++ b/arm_compute/graph/Graph.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -216,7 +216,7 @@
      *
      * @return Tensor ID
      */
-    TensorID create_tensor(TensorDescriptor desc = TensorDescriptor());
+    TensorID create_tensor(const TensorDescriptor &desc = TensorDescriptor());
 
 private:
     GraphID                              _id      = GraphID(0); /**< Graph id */
diff --git a/arm_compute/graph/GraphBuilder.h b/arm_compute/graph/GraphBuilder.h
index 1296f56..1d6ecc8 100644
--- a/arm_compute/graph/GraphBuilder.h
+++ b/arm_compute/graph/GraphBuilder.h
@@ -25,6 +25,7 @@
 #define __ARM_COMPUTE_GRAPH_GRAPH_BUILDER_H__
 
 #include "arm_compute/graph/ITensorAccessor.h"
+#include "arm_compute/graph/LayerDescriptors.h"
 #include "arm_compute/graph/Types.h"
 
 namespace arm_compute
@@ -50,7 +51,7 @@
      *
      * @return Node ID of the created node, EmptyNodeID in case of error
      */
-    static NodeID add_const_node(Graph &g, NodeParams params, TensorDescriptor desc, ITensorAccessorUPtr accessor = nullptr);
+    static NodeID add_const_node(Graph &g, NodeParams params, const TensorDescriptor &desc, ITensorAccessorUPtr accessor = nullptr);
     /** Adds an input layer node to the graph
      *
      * @param[in] g        Graph to add the node to
@@ -60,7 +61,7 @@
      *
      * @return Node ID of the created node, EmptyNodeID in case of error
      */
-    static NodeID add_input_node(Graph &g, NodeParams params, TensorDescriptor desc, ITensorAccessorUPtr accessor = nullptr);
+    static NodeID add_input_node(Graph &g, NodeParams params, const TensorDescriptor &desc, ITensorAccessorUPtr accessor = nullptr);
     /** Adds an output layer node to the graph
      *
      * @param[in] g        Graph to add the node to
@@ -73,14 +74,16 @@
     static NodeID add_output_node(Graph &g, NodeParams params, NodeIdxPair input, ITensorAccessorUPtr accessor = nullptr);
     /** Adds an activation layer node to the graph
      *
-     * @param[in] g        Graph to add the node to
-     * @param[in] params   Common node parameters
-     * @param[in] input    Input to the activation layer node as a NodeID-Index pair
-     * @param[in] act_info Activation layer information
+     * @param[in] g              Graph to add the node to
+     * @param[in] params         Common node parameters
+     * @param[in] input          Input to the activation layer node as a NodeID-Index pair
+     * @param[in] act_info       Activation layer information
+     * @param[in] out_quant_info (Optional) Output quantization info
      *
      * @return Node ID of the created node, EmptyNodeID in case of error
      */
-    static NodeID add_activation_node(Graph &g, NodeParams params, NodeIdxPair input, ActivationLayerInfo act_info);
+    static NodeID add_activation_node(Graph &g, NodeParams params, NodeIdxPair input, ActivationLayerInfo act_info,
+                                      const QuantizationInfo out_quant_info = QuantizationInfo());
     /** Adds a batch normalization layer node to the graph
      *
      * @param[in] g              Graph to add the node to
@@ -163,14 +166,14 @@
                                          ITensorAccessorUPtr weights_accessor = nullptr, ITensorAccessorUPtr bias_accessor = nullptr);
     /** Adds a depth concatenate node to the graph
      *
-     * @param[in] g      Graph to add the node to
-     * @param[in] params Common node parameters
-     * @param[in] inputs Inputs to the depth concatenate layer node as a NodeID-Index pair
-     * @param[in] axis   Concatenation axis
+     * @param[in] g                 Graph to add the node to
+     * @param[in] params            Common node parameters
+     * @param[in] inputs            Inputs to the concatenate layer node as a NodeID-Index pair
+     * @param[in] concat_descriptor Concatenation layer descriptor
      *
      * @return Node ID of the created node, EmptyNodeID in case of error
      */
-    static NodeID add_concatenate_node(Graph &g, NodeParams params, std::vector<NodeIdxPair> inputs, DataLayoutDimension axis);
+    static NodeID add_concatenate_node(Graph &g, NodeParams params, const std::vector<NodeIdxPair> &inputs, descriptors::ConcatLayerDescriptor concat_descriptor);
     /** Adds a depth-wise convolution layer node to the graph
      *
      * @param[in] g                     Graph to add the node to
@@ -183,13 +186,15 @@
      * @param[in] weights_accessor      (Optional) Accessor of the weights node data
      * @param[in] bias_accessor         (Optional) Accessor of the bias node data
      * @param[in] quant_info            (Optional) Weights quantization info
+     * @param[in] out_quant_info        (Optional) Output quantization info
      *
      * @return Node ID of the created node, EmptyNodeID in case of error
      */
     static NodeID add_depthwise_convolution_node(Graph &g, NodeParams params, NodeIdxPair input,
                                                  Size2D kernel_spatial_extend, PadStrideInfo conv_info, int depth_multiplier = 1,
                                                  DepthwiseConvolutionMethod method    = DepthwiseConvolutionMethod::Default,
-                                                 ITensorAccessorUPtr weights_accessor = nullptr, ITensorAccessorUPtr bias_accessor = nullptr, const QuantizationInfo quant_info = QuantizationInfo());
+                                                 ITensorAccessorUPtr weights_accessor = nullptr, ITensorAccessorUPtr bias_accessor = nullptr, const QuantizationInfo quant_info = QuantizationInfo(),
+                                                 const QuantizationInfo out_quant_info = QuantizationInfo());
     /** Adds an element-wise layer node to the graph
      *
      * @param[in] g         Graph to add the node to
@@ -212,7 +217,7 @@
      *
      * @return Node ID of the created node, EmptyNodeID in case of error
      */
-    static NodeID add_detection_output_node(Graph &g, NodeParams params, NodeIdxPair input_loc, NodeIdxPair input_conf, NodeIdxPair input_priorbox, DetectionOutputLayerInfo detect_info);
+    static NodeID add_detection_output_node(Graph &g, NodeParams params, NodeIdxPair input_loc, NodeIdxPair input_conf, NodeIdxPair input_priorbox, const DetectionOutputLayerInfo &detect_info);
     /** Adds a Dummy node to the graph
      *
      * @note this node if for debugging purposes. Just alters the shape of the graph pipeline as requested.
@@ -236,6 +241,23 @@
     static NodeID add_flatten_node(Graph &g, NodeParams params, NodeIdxPair input);
     /** Adds a fully connected layer node to the graph
      *
+     * @param[in] g              Graph to add the layer to
+     * @param[in] params         Common node parameters
+     * @param[in] input          Input to the fully connected layer node as a NodeID-Index pair
+     * @param[in] num_outputs    Number of output neurons
+     * @param[in] weights_nid    Node ID of the weights node data
+     * @param[in] bias_nid       (Optional) Node ID of the bias node data. Defaults to EmptyNodeID
+     * @param[in] fc_info        (Optional) Fully connected layer metadata
+     * @param[in] out_quant_info (Optional) Output quantization info
+     *
+     * @return Node ID of the created node, EmptyNodeID in case of error
+     */
+    static NodeID add_fully_connected_layer(Graph &g, NodeParams params, NodeIdxPair input, unsigned int num_outputs,
+                                            NodeID weights_nid, NodeID bias_nid = EmptyNodeID,
+                                            const FullyConnectedLayerInfo fc_info        = FullyConnectedLayerInfo(),
+                                            const QuantizationInfo        out_quant_info = QuantizationInfo());
+    /** Adds a fully connected layer node to the graph
+     *
      * @param[in] g                  Graph to add the layer to
      * @param[in] params             Common node parameters
      * @param[in] input              Input to the fully connected layer node as a NodeID-Index pair
@@ -331,7 +353,7 @@
      *
      * @return Node ID of the created node, EmptyNodeID in case of error
      */
-    static NodeID add_priorbox_node(Graph &g, NodeParams params, NodeIdxPair input0, NodeIdxPair input1, PriorBoxLayerInfo prior_info);
+    static NodeID add_priorbox_node(Graph &g, NodeParams params, NodeIdxPair input0, NodeIdxPair input1, const PriorBoxLayerInfo &prior_info);
     /** Adds a reorg layer node to the graph
      *
      * @param[in] g      Graph to add the node to
@@ -421,6 +443,16 @@
      * @return Node ID of the created node, EmptyNodeID in case of error
      */
     static NodeID add_split_node(Graph &g, NodeParams params, NodeIdxPair input, unsigned int num_splits, unsigned int axis = 0);
+    /** Adds a stack layer node to the graph
+     *
+     * @param[in] g      Graph to add the node to
+     * @param[in] params Common node parameters
+     * @param[in] inputs Inputs to the reorg layer node as a NodeID-Index pair
+     * @param[in] axis   Axis along which the input tensors have to be packed
+     *
+     * @return Node ID of the created node, EmptyNodeID in case of error
+     */
+    static NodeID add_stack_node(Graph &g, NodeParams params, const std::vector<NodeIdxPair> &inputs, int axis);
     /** Adds an upsample layer to the graph
      *
      * @param[in] g                 Graph to add the node to
diff --git a/arm_compute/graph/INode.h b/arm_compute/graph/INode.h
index 4219150..edff837 100644
--- a/arm_compute/graph/INode.h
+++ b/arm_compute/graph/INode.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -25,6 +25,7 @@
 #define __ARM_COMPUTE_GRAPH_INODE_H__
 
 #include "arm_compute/core/Error.h"
+#include "arm_compute/graph/LayerDescriptors.h"
 #include "arm_compute/graph/TensorDescriptor.h"
 #include "arm_compute/graph/Types.h"
 
diff --git a/arm_compute/graph/INodeVisitor.h b/arm_compute/graph/INodeVisitor.h
index 573d642..291fe7c 100644
--- a/arm_compute/graph/INodeVisitor.h
+++ b/arm_compute/graph/INodeVisitor.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -91,6 +91,11 @@
      * @param[in] n Node to visit.
      */
     virtual void visit(FullyConnectedLayerNode &n) = 0;
+    /** Visit FusedConvolutionBatchNormalizationNode.
+     *
+     * @param[in] n Node to visit.
+     */
+    virtual void visit(FusedConvolutionBatchNormalizationNode &n) = 0;
     /** Visit InputNode.
      *
      * @param[in] n Node to visit.
@@ -136,6 +141,11 @@
      * @param[in] n Node to visit.
      */
     virtual void visit(SplitLayerNode &n) = 0;
+    /** Visit StackLayerNode.
+     *
+     * @param[in] n Node to visit.
+     */
+    virtual void visit(StackLayerNode &n) = 0;
 };
 
 /** Default visitor implementation
@@ -195,6 +205,10 @@
     {
         default_visit();
     }
+    virtual void visit(FusedConvolutionBatchNormalizationNode &n) override
+    {
+        default_visit();
+    }
     virtual void visit(InputNode &n) override
     {
         default_visit();
@@ -231,6 +245,10 @@
     {
         default_visit();
     }
+    virtual void visit(StackLayerNode &n) override
+    {
+        default_visit();
+    }
 #endif /* DOXYGEN_SKIP_THIS */
 
     /** Function to be overloaded by the client and implement default behavior for the
diff --git a/arm_compute/graph/LayerDescriptors.h b/arm_compute/graph/LayerDescriptors.h
new file mode 100644
index 0000000..f52beab
--- /dev/null
+++ b/arm_compute/graph/LayerDescriptors.h
@@ -0,0 +1,69 @@
+/*
+ * Copyright (c) 2019 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_CONCAT_DESCRIPTOR_H__
+#define __ARM_COMPUTE_CONCAT_DESCRIPTOR_H__
+
+#include "arm_compute/core/Types.h"
+
+namespace arm_compute
+{
+namespace graph
+{
+namespace descriptors
+{
+/** Concatenate layer descriptor */
+struct ConcatLayerDescriptor
+{
+    /** Default constructor */
+    ConcatLayerDescriptor()
+        : axis(DataLayoutDimension::CHANNEL), output_qinfo()
+    {
+    }
+
+    /** Constructor concatenate layer descriptor
+     *
+     * @param[in] axis Axis.
+     */
+    ConcatLayerDescriptor(DataLayoutDimension axis)
+        : axis(axis), output_qinfo()
+    {
+    }
+
+    /** Constructor concatenate layer descriptor
+     *
+     * @param[in] axis         Axis.
+     * @param[in] output_qinfo Output quantization info.
+     */
+    ConcatLayerDescriptor(DataLayoutDimension axis, QuantizationInfo output_qinfo)
+        : axis(axis), output_qinfo(output_qinfo)
+    {
+    }
+
+    const DataLayoutDimension axis;         /**< Concatenation Axis */
+    const QuantizationInfo    output_qinfo; /**< Output quantizazion info */
+};
+} // namespace descriptor
+} // namespace graph
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_CONCAT_DESCRIPTOR_H__ */
\ No newline at end of file
diff --git a/arm_compute/graph/Tensor.h b/arm_compute/graph/Tensor.h
index 54fb258..07eec1e 100644
--- a/arm_compute/graph/Tensor.h
+++ b/arm_compute/graph/Tensor.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -108,7 +108,7 @@
      *
      * @return Bound edges
      */
-    const std::set<EdgeID> bound_edges() const;
+    std::set<EdgeID> bound_edges() const;
 
 private:
     TensorID                         _id;          /**< Tensor id */
diff --git a/arm_compute/graph/TypeLoader.h b/arm_compute/graph/TypeLoader.h
index 77f0961..41f382a 100644
--- a/arm_compute/graph/TypeLoader.h
+++ b/arm_compute/graph/TypeLoader.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -100,6 +100,53 @@
     target = target_from_name(value);
     return stream;
 }
+
+/** Converts a string to a strong types enumeration @ref ConvolutionMethod
+ *
+ * @param[in] name String to convert
+ *
+ * @return Converted Target enumeration
+ */
+ConvolutionMethod Convolution_method_from_name(const std::string &name);
+
+/** Input Stream operator for @ref ConvolutionMethod
+ *
+ * @param[in]  stream Stream to parse
+ * @param[out] target Output target
+ *
+ * @return Updated stream
+ */
+inline ::std::istream &operator>>(::std::istream &stream, ConvolutionMethod &target)
+{
+    std::string value;
+    stream >> value;
+    target = Convolution_method_from_name(value);
+    return stream;
+}
+
+/** Converts a string to a strong types enumeration @ref DepthwiseConvolutionMethod
+ *
+ * @param[in] name String to convert
+ *
+ * @return Converted Target enumeration
+ */
+DepthwiseConvolutionMethod depthwise_convolution_method_from_name(const std::string &name);
+
+/** Input Stream operator for @ref DepthwiseConvolutionMethod
+ *
+ * @param[in]  stream Stream to parse
+ * @param[out] target Output target
+ *
+ * @return Updated stream
+ */
+inline ::std::istream &operator>>(::std::istream &stream, DepthwiseConvolutionMethod &target)
+{
+    std::string value;
+    stream >> value;
+    target = depthwise_convolution_method_from_name(value);
+    return stream;
+}
+
 } // namespace graph
 } // namespace arm_compute
 #endif /* __ARM_COMPUTE_GRAPH_TYPE_LOADER_H__ */
diff --git a/arm_compute/graph/TypePrinter.h b/arm_compute/graph/TypePrinter.h
index ca62d4e..29a2981 100644
--- a/arm_compute/graph/TypePrinter.h
+++ b/arm_compute/graph/TypePrinter.h
@@ -98,6 +98,9 @@
         case NodeType::FullyConnectedLayer:
             os << "FullyConnectedLayer";
             break;
+        case NodeType::FusedConvolutionBatchNormalizationLayer:
+            os << "FusedConvolutionBatchNormalizationLayer";
+            break;
         case NodeType::GenerateProposalsLayer:
             os << "GenerateProposalsLayer";
             break;
@@ -140,6 +143,9 @@
         case NodeType::SplitLayer:
             os << "SplitLayer";
             break;
+        case NodeType::StackLayer:
+            os << "StackLayer";
+            break;
         case NodeType::UpsampleLayer:
             os << "UpsampleLayer";
             break;
diff --git a/arm_compute/graph/Types.h b/arm_compute/graph/Types.h
index 8377253..4d9e031 100644
--- a/arm_compute/graph/Types.h
+++ b/arm_compute/graph/Types.h
@@ -26,6 +26,7 @@
 
 #include "arm_compute/core/Error.h"
 #include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/CL/CLTunerTypes.h"
 
 #include <limits>
 #include <string>
@@ -34,6 +35,7 @@
 {
 namespace graph
 {
+using arm_compute::CLTunerMode;
 using arm_compute::Status;
 
 using arm_compute::Coordinates;
@@ -71,13 +73,13 @@
 
 // Forward declarations
 class TensorDescriptor;
-
 /** Graph configuration structure */
 struct GraphConfig
 {
     bool        use_function_memory_manager{ true };   /**< Use a memory manager to manage per-funcion auxilary memory */
     bool        use_transition_memory_manager{ true }; /**< Use a memory manager to manager transition buffer memory */
     bool        use_tuner{ false };                    /**< Use a tuner in tunable backends */
+    CLTunerMode tuner_mode{ CLTunerMode::EXHAUSTIVE }; /**< Tuner mode to be used by the CL tuner */
     int         num_threads{ -1 };                     /**< Number of threads to use (thread capable backends), if 0 the backend will auto-initialize, if -1 the backend will stay as it is. */
     std::string tuner_file{ "acl_tuner.csv" };         /**< File to load/store tuning values from */
 };
@@ -138,6 +140,7 @@
     EltwiseLayer,
     FlattenLayer,
     FullyConnectedLayer,
+    FusedConvolutionBatchNormalizationLayer,
     GenerateProposalsLayer,
     NormalizationLayer,
     NormalizePlanarYUVLayer,
@@ -152,6 +155,7 @@
     SoftmaxLayer,
     SliceLayer,
     SplitLayer,
+    StackLayer,
     UpsampleLayer,
     YOLOLayer,
 
diff --git a/arm_compute/graph/Utils.h b/arm_compute/graph/Utils.h
index 1a0509b..2fa2f3b 100644
--- a/arm_compute/graph/Utils.h
+++ b/arm_compute/graph/Utils.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -89,11 +89,12 @@
  * @return A PassManager with default mutating passes
  */
 PassManager create_default_pass_manager(Target target);
-/** Default setups the graph context if not done manually
+/** Setups requested backend context if it exists, is supported and hasn't been initialized already.
  *
- * @param[in,out] ctx Graph Context
+ * @param[in,out] ctx    Graph Context.
+ * @param[in]     target Target to setup the backend for.
  */
-void setup_default_graph_context(GraphContext &ctx);
+void setup_requested_backend_context(GraphContext &ctx, Target target);
 /** Default releases the graph context if not done manually
  *
  * @param[in,out] ctx Graph Context
@@ -109,12 +110,12 @@
 size_t get_dimension_size(const TensorDescriptor &descriptor, const DataLayoutDimension data_layout_dimension);
 /** Get index of a tensor's given dimension depending on its layout
  *
- * @param[in] descriptor            Descriptor
+ * @param[in] data_layout           Data layout of the tensor
  * @param[in] data_layout_dimension Tensor data layout dimension
  *
  * @return Idx of given dimension
  */
-size_t get_dimension_idx(const TensorDescriptor &descriptor, const DataLayoutDimension data_layout_dimension);
+size_t get_dimension_idx(DataLayout data_layout, const DataLayoutDimension data_layout_dimension);
 /** Get the list of driving nodes of a given node
  *
  * @param[in] node Node to find the driving node of
diff --git a/arm_compute/graph/backends/CL/CLDeviceBackend.h b/arm_compute/graph/backends/CL/CLDeviceBackend.h
index 49e7596..afe01ff 100644
--- a/arm_compute/graph/backends/CL/CLDeviceBackend.h
+++ b/arm_compute/graph/backends/CL/CLDeviceBackend.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -50,6 +50,11 @@
      * @param[in] enable_tuning Enables tuning if false else true
      */
     void set_kernel_tuning(bool enable_tuning);
+    /** Set kernel tuning mode
+     *
+     * @param[in] tuning_mode Indicates how exhaustive the search for the optimal LWS should be while tuning
+     */
+    void set_kernel_tuning_mode(CLTunerMode tuning_mode);
 
     // Inherited overridden methods
     void initialize_backend() override;
diff --git a/arm_compute/graph/backends/FunctionHelpers.h b/arm_compute/graph/backends/FunctionHelpers.h
index 7242bc6..f6e6286 100644
--- a/arm_compute/graph/backends/FunctionHelpers.h
+++ b/arm_compute/graph/backends/FunctionHelpers.h
@@ -28,6 +28,8 @@
 #include "arm_compute/graph/Tensor.h"
 #include "arm_compute/graph/TypePrinter.h"
 #include "arm_compute/graph/Types.h"
+#include "arm_compute/graph/Utils.h"
+#include "arm_compute/graph/backends/FusedConvolutionBatchNormalizationFunction.h"
 #include "arm_compute/graph/backends/Utils.h"
 #include "arm_compute/graph/nodes/Nodes.h"
 
@@ -108,7 +110,7 @@
     ARM_COMPUTE_LOG_GRAPH_INFO("Instantiated "
                                << node.name()
                                << " Type: " << node.type()
-                               << " Target " << TargetInfo::TargetType
+                               << " Target: " << TargetInfo::TargetType
                                << " Data Type: " << input->info()->data_type()
                                << " Shape: " << input->info()->tensor_shape()
                                << " Activation function: " << act_info.activation()
@@ -135,11 +137,12 @@
     validate_node<TargetInfo>(node, 5 /* expected inputs */, 1 /* expected outputs */);
 
     // Extract IO and info
-    typename TargetInfo::TensorType *input     = get_backing_tensor<TargetInfo>(node.input(0));
-    typename TargetInfo::TensorType *mean      = get_backing_tensor<TargetInfo>(node.input(1));
-    typename TargetInfo::TensorType *var       = get_backing_tensor<TargetInfo>(node.input(2));
-    typename TargetInfo::TensorType *beta      = get_backing_tensor<TargetInfo>(node.input(3));
-    typename TargetInfo::TensorType *gamma     = get_backing_tensor<TargetInfo>(node.input(4));
+    typename TargetInfo::TensorType *input = get_backing_tensor<TargetInfo>(node.input(0));
+    typename TargetInfo::TensorType *mean  = get_backing_tensor<TargetInfo>(node.input(1));
+    typename TargetInfo::TensorType *var   = get_backing_tensor<TargetInfo>(node.input(2));
+    typename TargetInfo::TensorType *beta  = get_backing_tensor<TargetInfo>(node.input(3));
+    typename TargetInfo::TensorType *gamma = get_backing_tensor<TargetInfo>(node.input(4));
+
     typename TargetInfo::TensorType *output    = get_backing_tensor<TargetInfo>(node.output(0));
     const float                      epsilon   = node.epsilon();
     const ActivationLayerInfo        fused_act = node.fused_activation();
@@ -163,6 +166,61 @@
     return std::move(func);
 }
 
+/** Create a backend batch normalization layer function
+ *
+ * @tparam BatchNormalizationLayerFunction Backend batch normalization function
+ * @tparam TargetInfo                      Target-specific information
+ *
+ * @param[in] node Node to create the backend function for
+ *
+ * @return Backend batch normalization layer function
+ */
+template <typename FusedLayerTypes, typename TargetInfo>
+std::unique_ptr<IFunction> create_fused_convolution_batch_normalization_layer(FusedConvolutionBatchNormalizationNode &node)
+{
+    validate_node<TargetInfo>(node, 7 /* expected inputs */, 1 /* expected outputs */);
+
+    // Extract IO and info
+    typename TargetInfo::TensorType *input   = get_backing_tensor<TargetInfo>(node.input(0));
+    typename TargetInfo::TensorType *weights = get_backing_tensor<TargetInfo>(node.input(1));
+    typename TargetInfo::TensorType *biases  = get_backing_tensor<TargetInfo>(node.input(2));
+    typename TargetInfo::TensorType *mean    = get_backing_tensor<TargetInfo>(node.input(3));
+    typename TargetInfo::TensorType *var     = get_backing_tensor<TargetInfo>(node.input(4));
+    typename TargetInfo::TensorType *beta    = get_backing_tensor<TargetInfo>(node.input(5));
+    typename TargetInfo::TensorType *gamma   = get_backing_tensor<TargetInfo>(node.input(6));
+
+    typename TargetInfo::TensorType *output = get_backing_tensor<TargetInfo>(node.output(0));
+
+    const PadStrideInfo       conv_info  = node.convolution_info();
+    const unsigned int        num_groups = node.num_groups();
+    const bool                fast_math  = node.fast_math_hint() == FastMathHint::Enabled;
+    const ActivationLayerInfo fused_act  = node.fused_activation();
+    const float               epsilon    = node.epsilon();
+
+    const bool is_quantized = is_data_type_quantized_asymmetric(input->info()->data_type());
+    if(is_quantized && biases != nullptr)
+    {
+        biases->info()->set_data_type(DataType::S32);
+    }
+
+    // Create and configure function
+    auto func = support::cpp14::make_unique<FusedConvolutionBatchNormalizationFunction<TargetInfo, FusedLayerTypes>>();
+    func->configure(input, weights, biases, output, mean, var, beta, gamma, epsilon, conv_info, num_groups, fast_math, fused_act);
+
+    // Log info
+    ARM_COMPUTE_LOG_GRAPH_INFO("Instantiated "
+                               << node.name()
+                               << " Type: " << node.name()
+                               << " Target: " << TargetInfo::TargetType
+                               << " Data Type: " << input->info()->data_type()
+                               << " Input shape: " << input->info()->tensor_shape()
+                               << " Weights shape: " << weights->info()->tensor_shape()
+                               << " Output shape: " << output->info()->tensor_shape()
+                               << (fused_act.enabled() ? " " + to_string(fused_act.activation()) : "")
+                               << std::endl);
+    return std::move(func);
+}
+
 /** Create a backend bounding box transform layer function
  *
  * @tparam BoundingBoxTransformLayerFunction    Backend bounding box transform function
@@ -188,8 +246,10 @@
     func->configure(input, output, deltas, bbox_info);
 
     // Log info
-    ARM_COMPUTE_LOG_GRAPH_INFO("Instantiated " << node.type()
-                               << " Target " << TargetInfo::TargetType
+    ARM_COMPUTE_LOG_GRAPH_INFO("Instantiated "
+                               << node.name()
+                               << " Type: " << node.type()
+                               << " Target: " << TargetInfo::TargetType
                                << " Data Type: " << input->info()->data_type()
                                << " Shape: " << input->info()->tensor_shape()
                                << " BoundingBox Info img W: " << bbox_info.img_width() << " "
@@ -262,13 +322,20 @@
         inputs.push_back(get_backing_tensor<TargetInfo>(node.input(i)));
     }
     typename TargetInfo::TensorType *output      = get_backing_tensor<TargetInfo>(node.output(0));
-    const DataLayoutDimension        concat_axis = node.concatenation_axis();
+    const DataLayout                 data_layout = node.output(0) != nullptr ? node.output(0)->desc().layout : DataLayout::UNKNOWN;
+    const size_t                     concat_axis = get_dimension_idx(data_layout, node.concatenation_axis());
 
     // Create and configure function
     auto func = support::cpp14::make_unique<ConcatenateLayerFunction>();
     func->configure(inputs, output, concat_axis);
 
     // Log info
+    const bool         is_quantized = is_data_type_quantized_asymmetric(output->info()->data_type());
+    std::ostringstream qss;
+    if(is_quantized)
+    {
+        qss << " Output QuantInfo: " << output->info()->quantization_info();
+    }
     ARM_COMPUTE_LOG_GRAPH_INFO("Instantiated "
                                << node.name()
                                << " Type: " << node.type()
@@ -277,6 +344,7 @@
                                << " Shape: " << output->info()->tensor_shape()
                                << " Num Inputs: " << inputs.size()
                                << " Axis: " << concat_axis
+                               << qss.str()
                                << std::endl);
 
     return std::move(func);
@@ -364,10 +432,10 @@
                                << " Target: " << TargetInfo::TargetType
                                << " Data Type: " << input->info()->data_type()
                                << " Groups: " << num_groups
-                               << qss.str()
                                << " Input shape: " << input->info()->tensor_shape()
                                << " Weights shape: " << weights->info()->tensor_shape()
                                << " Output shape: " << output->info()->tensor_shape()
+                               << qss.str()
                                << (fused_act.enabled() ? " " + to_string(fused_act.activation()) : "")
                                << std::endl);
     return func;
@@ -479,11 +547,11 @@
                                << " Type: " << func_name
                                << " Target: " << TargetInfo::TargetType
                                << " Data Type: " << input->info()->data_type()
-                               << qss.str()
                                << " Input shape: " << input->info()->tensor_shape()
                                << " Weights shape: " << weights->info()->tensor_shape()
                                << " Output shape: " << output->info()->tensor_shape()
                                << " Depth multiplier: " << depth_multiplier
+                               << qss.str()
                                << (fused_act.enabled() ? " " + to_string(fused_act.activation()) : "")
                                << std::endl);
     return func;
@@ -1120,8 +1188,10 @@
     func->configure(input, rois, output, pool_info);
 
     // Log info
-    ARM_COMPUTE_LOG_GRAPH_INFO("Instantiated " << node.type()
-                               << " Target " << TargetInfo::TargetType
+    ARM_COMPUTE_LOG_GRAPH_INFO("Instantiated "
+                               << node.name()
+                               << " Type: " << node.type()
+                               << " Target: " << TargetInfo::TargetType
                                << " Data Type: " << input->info()->data_type()
                                << " Input shape: " << input->info()->tensor_shape()
                                << " Output shape: " << output->info()->tensor_shape()
@@ -1208,6 +1278,49 @@
 
     return std::move(func);
 }
+
+/** Create a backend layer stack function
+ *
+ * @tparam StackLayerFunction Backend stack function
+ * @tparam TargetInfo         Target-specific information
+ *
+ * @param[in] node Node to create the backend function for
+ *
+ * @return Backend stack layer function
+ */
+template <typename StackLayerFunction, typename TargetInfo>
+std::unique_ptr<arm_compute::IFunction> create_stack_layer(StackLayerNode &node)
+{
+    ARM_COMPUTE_LOG_GRAPH_VERBOSE("Creating Stack node with ID : " << node.id() << " and Name: " << node.name() << std::endl);
+    ARM_COMPUTE_ERROR_ON(node.num_outputs() != 1);
+
+    // Extract IO and info
+    std::vector<typename TargetInfo::TensorType *> inputs;
+    for(unsigned int i = 0; i < node.num_inputs(); ++i)
+    {
+        inputs.push_back(get_backing_tensor<TargetInfo>(node.input(i)));
+    }
+    typename TargetInfo::TensorType *output = get_backing_tensor<TargetInfo>(node.output(0));
+    const int                        axis   = node.axis();
+
+    // Create and configure function
+    auto func = support::cpp14::make_unique<StackLayerFunction>();
+    func->configure(inputs, axis, output);
+
+    // Log info
+    ARM_COMPUTE_LOG_GRAPH_INFO("Instantiated "
+                               << node.name()
+                               << " Type: " << node.type()
+                               << " Target: " << TargetInfo::TargetType
+                               << " Data Type: " << output->info()->data_type()
+                               << " Inputs shape: " << inputs[0]->info()->tensor_shape()
+                               << " Output shape: " << output->info()->tensor_shape()
+                               << " Num Inputs: " << inputs.size()
+                               << " Axis: " << axis
+                               << std::endl);
+
+    return std::move(func);
+}
 /** Create a backend Upsample layer function
  *
  * @tparam UpsampleLayerFunction Backend Upsample function
diff --git a/arm_compute/graph/backends/FusedConvolutionBatchNormalizationFunction.h b/arm_compute/graph/backends/FusedConvolutionBatchNormalizationFunction.h
new file mode 100644
index 0000000..92af17b
--- /dev/null
+++ b/arm_compute/graph/backends/FusedConvolutionBatchNormalizationFunction.h
@@ -0,0 +1,133 @@
+/*
+ * Copyright (c) 2019 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_BACKENDS_FUSED_CONVOLUTION_BATCH_NORMAZLIZATION_FUNCTION_H__
+#define __ARM_COMPUTE_GRAPH_BACKENDS_FUSED_CONVOLUTION_BATCH_NORMAZLIZATION_FUNCTION_H__
+
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/IFunction.h"
+
+namespace arm_compute
+{
+namespace graph
+{
+namespace backends
+{
+/** Wrapper function to first apply {NE, CL}BatchNormalizationLayer on the weights and then run {NE, CL}ConvolutionLayer with the modified weights */
+template <typename TargetInfo, typename FusedLayerTypes>
+class FusedConvolutionBatchNormalizationFunction : public IFunction
+{
+public:
+    using TensorType         = typename TargetInfo::TensorType;
+    using TensorConcreteType = typename TargetInfo::TensorConcreteType;
+
+    FusedConvolutionBatchNormalizationFunction()
+        : _conv_layer(), _fused_batch_norm_layer(), _fused_bias(), _is_prepared(false)
+    {
+    }
+
+    /** Set the input and output tensors.
+     *
+     * @param[in]  input      Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
+     *                        while every optional dimension from 4 and above represent a batch of inputs.
+     *                        Data types supported: QASYMM8/F16/F32.
+     * @param[in]  weights    Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported: Same as @p input.
+     * @param[in]  bias       Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM].
+     *                        Data type supported: Should match @p input data type, except for input of QASYMM8 type where biases should be of S32 type.
+     * @param[out] output     Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
+     *                        Data types 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 info. 1 dimension with size equal to the feature maps [FM]. If not provided, default value for beta is 0. Data types supported: Same as @p input
+     * @param[in]  gamma      Gamma values tensor info. 1 dimension with size equal to the feature maps [FM]. If not provided, default value for gamma is 1. Data types supported: Same as @p input
+     * @param[in]  epsilon    Small value to avoid division with zero. Default value is 0.001f.
+     * @param[in]  conv_info  Contains padding and stride information described in @ref PadStrideInfo.
+     * @param[in]  num_groups Number of groups when performing a grouped convolution. num_groups != 1 is only supported for NCHW data layout
+     * @param[in]  fast_math  Enable fast math computation. In case this flag were set, the function could dispatch the fastest implementation
+     *                        available which may introduce a drop of accuracy as well. Default is false
+     * @param[in]  fused_act  Activation layer information in case of a fused activation.
+     *
+     */
+    void configure(TensorType       *input,
+                   TensorType       *weights,
+                   TensorType       *bias,
+                   TensorType       *output,
+                   const TensorType *mean,
+                   const TensorType *var,
+                   const TensorType *beta,
+                   const TensorType *gamma,
+                   float epsilon, const PadStrideInfo &conv_info, unsigned int num_groups, bool fast_math, ActivationLayerInfo const &fused_act)
+    {
+        // We don't run any validate, as we assume that the layers have been already validated
+        const bool        has_bias = (bias != nullptr);
+        const TensorType *bias_to_use;
+
+        // We check if the layer has a bias. If yes, use it in-place. If not, we need to create one
+        // as batch normalization might end up with a bias != 0
+        if(has_bias)
+        {
+            _fused_batch_norm_layer.configure(weights, mean, var, nullptr, nullptr, bias, beta, gamma, epsilon);
+            bias_to_use = bias;
+        }
+        else
+        {
+            _fused_batch_norm_layer.configure(weights, mean, var, nullptr, &_fused_bias, nullptr, beta, gamma, epsilon);
+            bias_to_use = &_fused_bias;
+        }
+
+        _conv_layer.configure(input, weights, bias_to_use, output, conv_info, WeightsInfo(), Size2D(1U, 1U), fused_act, fast_math, num_groups);
+
+        if(!has_bias)
+        {
+            _fused_bias.allocator()->allocate();
+        }
+    }
+
+    // Inherited methods overridden:
+    void run()
+    {
+        prepare();
+        _conv_layer.run();
+    }
+
+    void prepare()
+    {
+        if(!_is_prepared)
+        {
+            _fused_batch_norm_layer.run();
+            _is_prepared = true;
+        }
+    }
+
+private:
+    typename FusedLayerTypes::ConvolutionLayer       _conv_layer;
+    typename FusedLayerTypes::FuseBatchNormalization _fused_batch_norm_layer;
+    TensorConcreteType                               _fused_bias;
+    bool                                             _is_prepared;
+};
+} // namespace backends
+} // namespace graph
+} // namespace arm_compute
+
+#endif /* __ARM_COMPUTE_GRAPH_BACKENDS_FUSED_CONVOLUTION_BATCH_NORMAZLIZATION_FUNCTION_H__ */
diff --git a/arm_compute/graph/backends/ValidateHelpers.h b/arm_compute/graph/backends/ValidateHelpers.h
index 8942be2..dbf8f35 100644
--- a/arm_compute/graph/backends/ValidateHelpers.h
+++ b/arm_compute/graph/backends/ValidateHelpers.h
@@ -203,6 +203,7 @@
 
     return status;
 }
+
 /** Validates a detection output layer node
  *
  * @tparam DetectionOutputLayer DetectionOutput layer type
@@ -372,6 +373,29 @@
     return ReorgLayer::validate(input, output, node.stride());
 }
 
+/** Validates a Reshape layer node
+ *
+ * @tparam ReshapeLayer Reshape layer type
+ *
+ * @param[in] node Node to validate
+ *
+ * @return Status
+ */
+template <typename ReshapeLayer>
+Status validate_reshape_layer(ReshapeLayerNode &node)
+{
+    ARM_COMPUTE_LOG_GRAPH_VERBOSE("Validating ReshapeLayer node with ID : " << node.id() << " and Name: " << node.name() << std::endl);
+    ARM_COMPUTE_RETURN_ERROR_ON(node.num_inputs() != 1);
+    ARM_COMPUTE_RETURN_ERROR_ON(node.num_outputs() != 1);
+
+    // Extract input and output
+    arm_compute::ITensorInfo *input  = detail::get_backing_tensor_info(node.input(0));
+    arm_compute::ITensorInfo *output = detail::get_backing_tensor_info(node.output(0));
+
+    // Validate function
+    return ReshapeLayer::validate(input, output);
+}
+
 /** Validates a ROI Align layer node
  *
  * @tparam ROIAlignLayer ROIAlign layer type
diff --git a/arm_compute/graph/frontend/Layers.h b/arm_compute/graph/frontend/Layers.h
index 1a71c89..a4c03a6 100644
--- a/arm_compute/graph/frontend/Layers.h
+++ b/arm_compute/graph/frontend/Layers.h
@@ -72,22 +72,24 @@
 public:
     /** Construct an output layer.
      *
-     * @param[in] accessor Accessor to give output tensor data to.
+     * @param[in] accessor       Accessor to give output tensor data to.
+     * @param[in] connection_idx (Optional) Input connection index
      */
-    OutputLayer(ITensorAccessorUPtr accessor)
-        : _accessor(std::move(accessor))
+    OutputLayer(ITensorAccessorUPtr accessor, unsigned int connection_idx = 0)
+        : _accessor(std::move(accessor)), _connection_idx(connection_idx)
     {
     }
 
     NodeID create_layer(IStream &s) override
     {
         NodeParams  common_params = { name(), s.hints().target_hint };
-        NodeIdxPair input         = { s.tail_node(), 0 };
+        NodeIdxPair input         = { s.tail_node(), _connection_idx };
         return GraphBuilder::add_output_node(s.graph(), common_params, input, std::move(_accessor));
     }
 
 private:
     ITensorAccessorUPtr _accessor;
+    unsigned int        _connection_idx;
 };
 
 /** Activation Layer */
@@ -96,10 +98,13 @@
 public:
     /** Construct an activation layer.
      *
-     * @param[in] act_info Activation information
+     * @param[in] act_info       Activation information
+     * @param[in] out_quant_info (Optional) Output quantization info
      */
-    ActivationLayer(ActivationLayerInfo act_info)
-        : _act_info(act_info)
+    ActivationLayer(ActivationLayerInfo    act_info,
+                    const QuantizationInfo out_quant_info = QuantizationInfo())
+        : _act_info(act_info),
+          _out_quant_info(std::move(out_quant_info))
     {
     }
 
@@ -107,11 +112,12 @@
     {
         NodeParams  common_params = { name(), s.hints().target_hint };
         NodeIdxPair input         = { s.tail_node(), 0 };
-        return GraphBuilder::add_activation_node(s.graph(), common_params, input, _act_info);
+        return GraphBuilder::add_activation_node(s.graph(), common_params, input, _act_info, std::move(_out_quant_info));
     }
 
 private:
-    ActivationLayerInfo _act_info;
+    ActivationLayerInfo    _act_info;
+    const QuantizationInfo _out_quant_info;
 };
 
 /** Batchnormalization Layer */
@@ -225,7 +231,7 @@
      */
     template <typename... Ts>
     ConcatLayer(SubStream &&sub_stream1, SubStream &&sub_stream2, Ts &&... rest_sub_streams)
-        : _sub_streams(), _axis(DataLayoutDimension::CHANNEL)
+        : _sub_streams(), _concat_descriptor(DataLayoutDimension::CHANNEL)
     {
         _sub_streams.push_back(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream1)));
         _sub_streams.push_back(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream2)));
@@ -238,14 +244,14 @@
     }
     /** Construct a concatenation layer
      *
-     * @param[in] axis             Axis over the concatenation will be performed
-     * @param[in] sub_stream1      First graph branch
-     * @param[in] sub_stream2      Second graph branch
-     * @param[in] rest_sub_streams Rest sub-graph branches
+     * @param[in] concat_descriptor Concat layer descriptor
+     * @param[in] sub_stream1       First graph branch
+     * @param[in] sub_stream2       Second graph branch
+     * @param[in] rest_sub_streams  Rest sub-graph branches
      */
     template <typename... Ts>
-    ConcatLayer(DataLayoutDimension axis, SubStream &&sub_stream1, SubStream &&sub_stream2, Ts &&... rest_sub_streams)
-        : _sub_streams(), _axis(axis)
+    ConcatLayer(descriptors::ConcatLayerDescriptor concat_descriptor, SubStream &&sub_stream1, SubStream &&sub_stream2, Ts &&... rest_sub_streams)
+        : _sub_streams(), _concat_descriptor(concat_descriptor)
     {
         _sub_streams.push_back(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream1)));
         _sub_streams.push_back(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream2)));
@@ -262,7 +268,7 @@
      */
     template <typename... Ts>
     ConcatLayer(SubStream &&sub_stream)
-        : _sub_streams(), _axis(DataLayoutDimension::CHANNEL)
+        : _sub_streams(), _concat_descriptor(DataLayoutDimension::CHANNEL)
     {
         _sub_streams.push_back(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream)));
     }
@@ -289,14 +295,14 @@
                     }
                 }
             }
-            nid = GraphBuilder::add_concatenate_node(s.graph(), common_params, nodes, _axis);
+            nid = GraphBuilder::add_concatenate_node(s.graph(), common_params, nodes, _concat_descriptor);
         }
         return nid;
     }
 
 private:
     std::vector<std::unique_ptr<SubStream>> _sub_streams;
-    DataLayoutDimension                     _axis;
+    descriptors::ConcatLayerDescriptor      _concat_descriptor;
 };
 
 /** Convolution Layer */
@@ -414,28 +420,31 @@
 public:
     /** Construct a depthwise convolution layer.
      *
-     * @param[in] conv_width       Convolution width.
-     * @param[in] conv_height      Convolution height.
-     * @param[in] weights          Accessor to get kernel weights from.
-     * @param[in] bias             Accessor to get kernel bias from.
-     * @param[in] conv_info        Padding and stride information.
-     * @param[in] depth_multiplier (Optional) Depth multiplier parameter.
-     * @param[in] quant_info       (Optional) Quantization info used for weights
+     * @param[in] conv_width         Convolution width.
+     * @param[in] conv_height        Convolution height.
+     * @param[in] weights            Accessor to get kernel weights from.
+     * @param[in] bias               Accessor to get kernel bias from.
+     * @param[in] conv_info          Padding and stride information.
+     * @param[in] depth_multiplier   (Optional) Depth multiplier parameter.
+     * @param[in] weights_quant_info (Optional) Quantization info used for weights
+     * @param[in] out_quant_info     (Optional) Output quantization info
      */
     DepthwiseConvolutionLayer(unsigned int           conv_width,
                               unsigned int           conv_height,
                               ITensorAccessorUPtr    weights,
                               ITensorAccessorUPtr    bias,
                               PadStrideInfo          conv_info,
-                              int                    depth_multiplier = 1,
-                              const QuantizationInfo quant_info       = QuantizationInfo())
+                              int                    depth_multiplier   = 1,
+                              const QuantizationInfo weights_quant_info = QuantizationInfo(),
+                              const QuantizationInfo out_quant_info     = QuantizationInfo())
         : _conv_width(conv_width),
           _conv_height(conv_height),
           _conv_info(std::move(conv_info)),
           _weights(std::move(weights)),
           _bias(std::move(bias)),
           _depth_multiplier(depth_multiplier),
-          _quant_info(std::move(quant_info))
+          _weights_quant_info(std::move(weights_quant_info)),
+          _out_quant_info(std::move(out_quant_info))
     {
     }
 
@@ -446,7 +455,7 @@
         return GraphBuilder::add_depthwise_convolution_node(s.graph(), common_params,
                                                             input, Size2D(_conv_width, _conv_height), _conv_info, _depth_multiplier,
                                                             s.hints().depthwise_convolution_method_hint,
-                                                            std::move(_weights), std::move(_bias), std::move(_quant_info));
+                                                            std::move(_weights), std::move(_bias), std::move(_weights_quant_info), std::move(_out_quant_info));
     }
 
 private:
@@ -456,7 +465,8 @@
     ITensorAccessorUPtr    _weights;
     ITensorAccessorUPtr    _bias;
     int                    _depth_multiplier;
-    const QuantizationInfo _quant_info;
+    const QuantizationInfo _weights_quant_info;
+    const QuantizationInfo _out_quant_info;
 };
 /** DetectionOutput Layer */
 class DetectionOutputLayer final : public ILayer
@@ -468,7 +478,7 @@
      * @param[in] sub_stream_prior PriorBox graph sub-stream.
      * @param[in] detect_info      DetectionOutput parameters.
      */
-    DetectionOutputLayer(SubStream &&sub_stream_conf, SubStream &&sub_stream_prior, DetectionOutputLayerInfo detect_info)
+    DetectionOutputLayer(SubStream &&sub_stream_conf, SubStream &&sub_stream_prior, const DetectionOutputLayerInfo &detect_info)
         : _ss_conf(std::move(sub_stream_conf)), _ss_prior(std::move(sub_stream_prior)), _detect_info(detect_info)
     {
     }
@@ -578,6 +588,34 @@
         : _num_outputs(num_outputs),
           _weights(std::move(weights)),
           _bias(std::move(bias)),
+          _weights_ss(nullptr),
+          _bias_ss(nullptr),
+          _fc_info(fc_info),
+          _weights_quant_info(std::move(weights_quant_info)),
+          _out_quant_info(std::move(out_quant_info))
+    {
+    }
+
+    /** Construct a fully connected layer.
+     *
+     * @param[in] num_outputs        Number of outputs.
+     * @param[in] sub_stream_weights Graph sub-stream for the weights.
+     * @param[in] sub_stream_bias    Graph sub-stream for the bias.
+     * @param[in] fc_info            (Optional) Fully connected layer metadata
+     * @param[in] weights_quant_info (Optional) Weights quantization information
+     * @param[in] out_quant_info     (Optional) Output quantization info
+     */
+    FullyConnectedLayer(unsigned int                  num_outputs,
+                        SubStream                   &&sub_stream_weights,
+                        SubStream                   &&sub_stream_bias,
+                        const FullyConnectedLayerInfo fc_info            = FullyConnectedLayerInfo(),
+                        const QuantizationInfo        weights_quant_info = QuantizationInfo(),
+                        const QuantizationInfo        out_quant_info     = QuantizationInfo())
+        : _num_outputs(num_outputs),
+          _weights(nullptr),
+          _bias(nullptr),
+          _weights_ss(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream_weights))),
+          _bias_ss(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream_bias))),
           _fc_info(fc_info),
           _weights_quant_info(std::move(weights_quant_info)),
           _out_quant_info(std::move(out_quant_info))
@@ -594,15 +632,29 @@
     {
         NodeParams  common_params = { name(), s.hints().target_hint };
         NodeIdxPair input         = { s.tail_node(), 0 };
-        return GraphBuilder::add_fully_connected_layer(s.graph(), common_params, input, _num_outputs,
-                                                       std::move(_weights), std::move(_bias), _fc_info,
-                                                       std::move(_weights_quant_info), std::move(_out_quant_info));
+        if(_weights != nullptr)
+        {
+            return GraphBuilder::add_fully_connected_layer(s.graph(), common_params, input, _num_outputs,
+                                                           std::move(_weights), std::move(_bias), _fc_info,
+                                                           std::move(_weights_quant_info), std::move(_out_quant_info));
+        }
+        else
+        {
+            ARM_COMPUTE_ERROR_ON(_weights_ss == nullptr);
+
+            NodeID bias_nid = (_bias_ss == nullptr) ? EmptyNodeID : _bias_ss->tail_node();
+            return GraphBuilder::add_fully_connected_layer(s.graph(), common_params, input, _num_outputs,
+                                                           _weights_ss->tail_node(), bias_nid, _fc_info,
+                                                           std::move(_out_quant_info));
+        }
     }
 
 private:
     unsigned int                  _num_outputs;
     ITensorAccessorUPtr           _weights;
     ITensorAccessorUPtr           _bias;
+    std::unique_ptr<SubStream>    _weights_ss;
+    std::unique_ptr<SubStream>    _bias_ss;
     const FullyConnectedLayerInfo _fc_info;
     const QuantizationInfo        _weights_quant_info;
     const QuantizationInfo        _out_quant_info;
@@ -786,7 +838,7 @@
      * @param[in] sub_stream First graph sub-stream
      * @param[in] prior_info PriorBox parameters.
      */
-    PriorBoxLayer(SubStream &&sub_stream, PriorBoxLayerInfo prior_info)
+    PriorBoxLayer(SubStream &&sub_stream, const PriorBoxLayerInfo &prior_info)
         : _ss(std::move(sub_stream)), _prior_info(prior_info)
     {
     }
@@ -986,6 +1038,92 @@
     float _beta;
 };
 
+/** Stack Layer */
+class StackLayer final : public ILayer
+{
+public:
+    /** Construct a concatenation layer
+     *
+     * @param[in] sub_stream1      First graph branch
+     * @param[in] sub_stream2      Second graph branch
+     * @param[in] rest_sub_streams Rest sub-graph branches
+     */
+    template <typename... Ts>
+    StackLayer(SubStream &&sub_stream1, SubStream &&sub_stream2, Ts &&... rest_sub_streams)
+        : _sub_streams(), _axis(0)
+    {
+        _sub_streams.push_back(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream1)));
+        _sub_streams.push_back(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream2)));
+
+        utility::for_each([&](SubStream && sub_stream)
+        {
+            _sub_streams.push_back(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream)));
+        },
+        std::move(rest_sub_streams)...);
+    }
+    /** Construct a concatenation layer
+     *
+     * @param[in] axis             Stack layer axis along which to stack the inputs
+     * @param[in] sub_stream1      First graph branch
+     * @param[in] sub_stream2      Second graph branch
+     * @param[in] rest_sub_streams Rest sub-graph branches
+     */
+    template <typename... Ts>
+    StackLayer(int axis, SubStream &&sub_stream1, SubStream &&sub_stream2, Ts &&... rest_sub_streams)
+        : _sub_streams(), _axis(axis)
+    {
+        _sub_streams.push_back(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream1)));
+        _sub_streams.push_back(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream2)));
+
+        utility::for_each([&](SubStream && sub_stream)
+        {
+            _sub_streams.push_back(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream)));
+        },
+        std::move(rest_sub_streams)...);
+    }
+    /** Construct a concat layer
+     *
+     * @param[in] sub_stream Sub-stream
+     */
+    template <typename... Ts>
+    StackLayer(SubStream &&sub_stream)
+        : _sub_streams(), _axis(0)
+    {
+        _sub_streams.push_back(arm_compute::support::cpp14::make_unique<SubStream>(std::move(sub_stream)));
+    }
+    NodeID create_layer(IStream &s) override
+    {
+        NodeID     nid           = EmptyNodeID;
+        NodeParams common_params = { name(), s.hints().target_hint };
+        if(_sub_streams.size() == 1 && _sub_streams.at(0) != nullptr)
+        {
+            nid = _sub_streams[0]->tail_node();
+        }
+        else
+        {
+            // Collect tail nodes and stack
+            std::vector<NodeIdxPair> nodes;
+            for(auto &ss : _sub_streams)
+            {
+                if(ss && (ss->tail_node() != EmptyNodeID))
+                {
+                    const auto tail_node = s.graph().node(ss->tail_node());
+                    if(tail_node != nullptr && tail_node->type() != NodeType::Output)
+                    {
+                        nodes.push_back({ ss->tail_node(), 0 });
+                    }
+                }
+            }
+            nid = GraphBuilder::add_stack_node(s.graph(), common_params, nodes, _axis);
+        }
+        return nid;
+    }
+
+private:
+    std::vector<std::unique_ptr<SubStream>> _sub_streams;
+    int                                     _axis;
+};
+
 /** Upsample Layer */
 class UpsampleLayer final : public ILayer
 {
diff --git a/arm_compute/graph/mutators/NodeFusionMutator.h b/arm_compute/graph/mutators/NodeFusionMutator.h
index 8f16c65..b9ca464 100644
--- a/arm_compute/graph/mutators/NodeFusionMutator.h
+++ b/arm_compute/graph/mutators/NodeFusionMutator.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -24,21 +24,13 @@
 #ifndef __ARM_COMPUTE_GRAPH_NODE_FUSION_MUTATOR_H__
 #define __ARM_COMPUTE_GRAPH_NODE_FUSION_MUTATOR_H__
 
+#include "arm_compute/graph/Graph.h"
 #include "arm_compute/graph/IGraphMutator.h"
 
 namespace arm_compute
 {
 namespace graph
 {
-namespace detail
-{
-/** Fused batch normalization with activation
- *
- * @param[in] g Graph to perform operation fusion on
- */
-void fuse_batch_norm_with_activation(Graph &g);
-} // namespace detail
-
 /** Mutation pass to fuss nodes */
 class NodeFusionMutator final : public IGraphMutator
 {
diff --git a/arm_compute/graph/nodes/ActivationLayerNode.h b/arm_compute/graph/nodes/ActivationLayerNode.h
index 570351b..a17b010 100644
--- a/arm_compute/graph/nodes/ActivationLayerNode.h
+++ b/arm_compute/graph/nodes/ActivationLayerNode.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -36,9 +36,11 @@
 public:
     /** Constructor
      *
-     * @param[in] info Activation Layer information
+     * @param[in] info           Activation Layer information
+     * @param[in] out_quant_info (Optional) Output quantization info
      */
-    ActivationLayerNode(ActivationLayerInfo info);
+    ActivationLayerNode(ActivationLayerInfo info,
+                        QuantizationInfo    out_quant_info = QuantizationInfo());
     /** Activation metadata accessor
      *
      * @return The activation info of the layer
@@ -51,8 +53,12 @@
     TensorDescriptor configure_output(size_t idx) const override;
     void accept(INodeVisitor &v) override;
 
+public:
+    static constexpr NodeType node_type = NodeType::ActivationLayer;
+
 private:
     ActivationLayerInfo _info;
+    QuantizationInfo    _out_quant_info;
 };
 } // namespace graph
 } // namespace arm_compute
diff --git a/arm_compute/graph/nodes/ConcatenateLayerNode.h b/arm_compute/graph/nodes/ConcatenateLayerNode.h
index 20c8523..fc12284 100644
--- a/arm_compute/graph/nodes/ConcatenateLayerNode.h
+++ b/arm_compute/graph/nodes/ConcatenateLayerNode.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -36,10 +36,10 @@
 public:
     /** Constructor
      *
-     * @param[in] total_nodes Number of nodes that will get concatenated
-     * @param[in] axis        Concatenation axis
+     * @param[in] total_nodes       Number of nodes that will get concatenated
+     * @param[in] concat_descriptor Concatenate Layer Descriptor
      */
-    ConcatenateLayerNode(unsigned int total_nodes, DataLayoutDimension axis);
+    ConcatenateLayerNode(unsigned int total_nodes, descriptors::ConcatLayerDescriptor concat_descriptor);
     /** Computes concatenations output descriptor
      *
      * @param[in] input_descriptors Input descriptors
@@ -68,6 +68,12 @@
      */
     DataLayoutDimension concatenation_axis() const;
 
+    /** Concatenation output quantization info accessor
+     *
+     * @return Output quantization info
+     */
+    QuantizationInfo output_quantization_info() const;
+
     // Inherited overridden methods:
     NodeType         type() const override;
     bool             forward_descriptors() override;
@@ -75,9 +81,9 @@
     void accept(INodeVisitor &v) override;
 
 private:
-    unsigned int        _total_nodes;
-    DataLayoutDimension _axis;
-    bool                _is_enabled;
+    unsigned int                       _total_nodes;
+    descriptors::ConcatLayerDescriptor _concat_descriptor;
+    bool                               _is_enabled;
 };
 } // namespace graph
 } // namespace arm_compute
diff --git a/arm_compute/graph/nodes/DepthwiseConvolutionLayerNode.h b/arm_compute/graph/nodes/DepthwiseConvolutionLayerNode.h
index 8c0aae1..fd02734 100644
--- a/arm_compute/graph/nodes/DepthwiseConvolutionLayerNode.h
+++ b/arm_compute/graph/nodes/DepthwiseConvolutionLayerNode.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -39,10 +39,12 @@
      * @param[in] info             Convolution layer attributes
      * @param[in] depth_multiplier (Optional) Depth multiplier parameter.
      * @param[in] method           (Optional) Depthwise convolution method to use
+     * @param[in] out_quant_info   (Optional) Output quantization info
      */
     DepthwiseConvolutionLayerNode(PadStrideInfo              info,
                                   int                        depth_multiplier = 1,
-                                  DepthwiseConvolutionMethod method           = DepthwiseConvolutionMethod::Default);
+                                  DepthwiseConvolutionMethod method           = DepthwiseConvolutionMethod::Default,
+                                  QuantizationInfo           out_quant_info   = QuantizationInfo());
     /** Sets the depthwise convolution method to use
      *
      * @param[in] method Depthwise convolution method to use
@@ -103,6 +105,7 @@
     PadStrideInfo              _info;
     int                        _depth_multiplier;
     DepthwiseConvolutionMethod _method;
+    QuantizationInfo           _out_quant_info;
     ActivationLayerInfo        _fused_activation;
 };
 } // namespace graph
diff --git a/arm_compute/graph/nodes/FusedConvolutionBatchNormalizationNode.h b/arm_compute/graph/nodes/FusedConvolutionBatchNormalizationNode.h
new file mode 100644
index 0000000..9b0f5b7
--- /dev/null
+++ b/arm_compute/graph/nodes/FusedConvolutionBatchNormalizationNode.h
@@ -0,0 +1,144 @@
+/*
+ * Copyright (c) 2019 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_FUSED_CONVOLUTION_BATCH_NORMALIZATION_NODE_H__
+#define __ARM_COMPUTE_GRAPH_FUSED_CONVOLUTION_BATCH_NORMALIZATION_NODE_H__
+
+#include "arm_compute/graph/INode.h"
+
+namespace arm_compute
+{
+namespace graph
+{
+/** Batch Normalization node */
+class FusedConvolutionBatchNormalizationNode final : public INode
+{
+public:
+    /** Constructor
+     *
+     * @param[in] epsilon          Epsilon parameter.
+     * @param[in] info             Convolution layer attributes.
+     * @param[in] num_groups       (Optional) Number of groups (Defaults to 1)
+     * @param[in] method           (Optional) Convolution method to use
+     * @param[in] fast_math_hint   (Optional) Fast math hint
+     * @param[in] out_quant_info   (Optional) Output quantization info
+     * @param[in] fused_activation (Optional) Fused activation layer. Disabled if not specified
+     */
+    FusedConvolutionBatchNormalizationNode(float epsilon, PadStrideInfo info,
+                                           unsigned int      num_groups     = 1,
+                                           ConvolutionMethod method         = ConvolutionMethod::Default,
+                                           FastMathHint      fast_math_hint = FastMathHint::Disabled,
+                                           QuantizationInfo out_quant_info = QuantizationInfo(), ActivationLayerInfo fused_activation = ActivationLayerInfo());
+
+    /** Epsilon parameter accessor
+     *
+     * @return Epsilon parameter
+     */
+    float epsilon() const;
+
+    /** Returns fused activation
+     *
+     * @return Fused activation
+     */
+    ActivationLayerInfo fused_activation() const;
+
+    /** Sets fused activation
+     *
+     * @param[in] fused_activation Fused activation to set
+     */
+    void set_fused_activation(ActivationLayerInfo fused_activation);
+
+    /** Computes convolution output descriptor
+     *
+     * @param[in] input_descriptor   Input descriptor
+     * @param[in] weights_descriptor Weights descriptor
+     * @param[in] info               Convolution operation attributes
+     *
+     * @return Output descriptor
+     */
+    static TensorDescriptor compute_output_descriptor(const TensorDescriptor &input_descriptor,
+                                                      const TensorDescriptor &weights_descriptor,
+                                                      const PadStrideInfo    &info);
+
+    /** Sets the convolution layer method to use
+     *
+     * @param[in] method Method to use for convolution
+     */
+    void set_convolution_method(ConvolutionMethod method);
+
+    /** Number of groups in convolution accessor
+     *
+     * @return Number of groups in convolution
+     */
+    unsigned int num_groups() const;
+
+    /** Convolution layer method accessor
+     *
+     * @note This is an indication on which convolution layer implementation to use,
+     *       if it fails to be created the library's heuristic approach will be used
+     *
+     * @return Convolution layer method to be used by the node
+     */
+    ConvolutionMethod convolution_method() const;
+
+    /** Sets the fast math fast hint
+     *
+     * @param[in] hint Hint to use for convolution
+     */
+    void set_fast_math_hint(FastMathHint hint);
+
+    /** Fast math hint accessor
+     *
+     * @return Fast math hint to be used by the node
+     */
+    FastMathHint fast_math_hint() const;
+
+    /** Convolution metadata accessor
+     *
+     * @return Convolution information
+     */
+    PadStrideInfo convolution_info() const;
+
+    // Inherited overridden methods:
+    NodeType         type() const override;
+    bool             forward_descriptors() override;
+    TensorDescriptor configure_output(size_t idx) const override;
+    void accept(INodeVisitor &v) override;
+
+public:
+    static constexpr NodeType node_type = NodeType::FusedConvolutionBatchNormalizationLayer;
+
+private:
+    float _epsilon;
+
+    PadStrideInfo       _info;
+    unsigned int        _num_groups;
+    ConvolutionMethod   _method;
+    FastMathHint        _fast_math_hint;
+    QuantizationInfo    _out_quant_info;
+    ActivationLayerInfo _fused_activation;
+};
+
+} // namespace graph
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_GRAPH_BATCH_NORMALIZATION_LAYER_NODE_H__ */
diff --git a/arm_compute/graph/nodes/Nodes.h b/arm_compute/graph/nodes/Nodes.h
index 2406485..4eb6a0f 100644
--- a/arm_compute/graph/nodes/Nodes.h
+++ b/arm_compute/graph/nodes/Nodes.h
@@ -38,6 +38,7 @@
 #include "arm_compute/graph/nodes/EltwiseLayerNode.h"
 #include "arm_compute/graph/nodes/FlattenLayerNode.h"
 #include "arm_compute/graph/nodes/FullyConnectedLayerNode.h"
+#include "arm_compute/graph/nodes/FusedConvolutionBatchNormalizationNode.h"
 #include "arm_compute/graph/nodes/GenerateProposalsLayerNode.h"
 #include "arm_compute/graph/nodes/InputNode.h"
 #include "arm_compute/graph/nodes/NormalizationLayerNode.h"
@@ -54,6 +55,7 @@
 #include "arm_compute/graph/nodes/SliceLayerNode.h"
 #include "arm_compute/graph/nodes/SoftmaxLayerNode.h"
 #include "arm_compute/graph/nodes/SplitLayerNode.h"
+#include "arm_compute/graph/nodes/StackLayerNode.h"
 #include "arm_compute/graph/nodes/UpsampleLayerNode.h"
 #include "arm_compute/graph/nodes/YOLOLayerNode.h"
 
diff --git a/arm_compute/graph/nodes/NodesFwd.h b/arm_compute/graph/nodes/NodesFwd.h
index cbda309..06c2e1f 100644
--- a/arm_compute/graph/nodes/NodesFwd.h
+++ b/arm_compute/graph/nodes/NodesFwd.h
@@ -44,6 +44,7 @@
 class EltwiseLayerNode;
 class FlattenLayerNode;
 class FullyConnectedLayerNode;
+class FusedConvolutionBatchNormalizationNode;
 class GenerateProposalsLayerNode;
 class InputNode;
 class NormalizationLayerNode;
@@ -60,6 +61,7 @@
 class SoftmaxLayerNode;
 class SliceLayerNode;
 class SplitLayerNode;
+class StackLayerNode;
 class UpsampleLayerNode;
 class YOLOLayerNode;
 } // namespace graph
diff --git a/arm_compute/graph/nodes/StackLayerNode.h b/arm_compute/graph/nodes/StackLayerNode.h
new file mode 100644
index 0000000..6c83fe2
--- /dev/null
+++ b/arm_compute/graph/nodes/StackLayerNode.h
@@ -0,0 +1,69 @@
+/*
+ * Copyright (c) 2019 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_STACK_LAYER_NODE_H__
+#define __ARM_COMPUTE_GRAPH_STACK_LAYER_NODE_H__
+
+#include "arm_compute/graph/INode.h"
+
+namespace arm_compute
+{
+namespace graph
+{
+/** Stack Layer node */
+class StackLayerNode final : public INode
+{
+public:
+    /** Constructor
+     *
+     * @param[in] total_nodes Number of nodes that will get stacked
+     * @param[in] axis        Axis alogn which to stack the input tensors
+     */
+    StackLayerNode(unsigned int total_nodes, int axis);
+    /** Computes stack output descriptor
+     *
+     * @param[in] input_descriptors Input descriptors
+     * @param[in] axis              Axis along which to stack the input tensors
+     *
+     * @return Expected output descriptor
+     */
+    static TensorDescriptor compute_output_descriptor(const std::vector<TensorDescriptor> &input_descriptors, int axis);
+    /** Stack axis parameter accessor
+     *
+     * @return Stack axis
+     */
+    int axis() const;
+
+    // Inherited overridden methods:
+    NodeType         type() const override;
+    bool             forward_descriptors() override;
+    TensorDescriptor configure_output(size_t idx) const override;
+    void accept(INodeVisitor &v) override;
+
+private:
+    unsigned int _total_nodes;
+    int          _axis;
+};
+} // namespace graph
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_GRAPH_STACK_LAYER_NODE_H__ */
diff --git a/arm_compute/graph/printers/DotGraphPrinter.h b/arm_compute/graph/printers/DotGraphPrinter.h
index d4cf692..9d2ea46 100644
--- a/arm_compute/graph/printers/DotGraphPrinter.h
+++ b/arm_compute/graph/printers/DotGraphPrinter.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -56,6 +56,7 @@
     void visit(ConvolutionLayerNode &n) override;
     void visit(DepthwiseConvolutionLayerNode &n) override;
     void visit(EltwiseLayerNode &n) override;
+    void visit(FusedConvolutionBatchNormalizationNode &n) override;
     void visit(NormalizationLayerNode &n) override;
     void visit(PoolingLayerNode &n) override;
     void default_visit() override;