IVGCVSW-7879 Change REVERSE_V2 from LayerWithParameters with 1 input, to Layer with 2 inputs

  * Changing ReverseV2 to use two inputs
  * This is required by the backends
  * The ReverseV2Descriptor was removed
  * Tests updated
  * Added a Run<> templatefor inputs with different data types

Signed-off-by: Tracy Narine <tracy.narine@arm.com>
Change-Id: I22f947de829b4b3da6bda3a74f4ffdef4052cc25

28 files changed, 665 insertions, 443 deletions

diff --git a/include/armnn/BackendHelper.hpp b/include/armnn/BackendHelper.hpp
index 6f804cbbed..6181ba5c40 100644
--- a/include/armnn/BackendHelper.hpp
+++ b/include/armnn/BackendHelper.hpp
@@ -360,9 +360,9 @@ public:
                            const ResizeDescriptor& descriptor,
                            Optional<std::string&> reasonIfUnsupported = EmptyOptional());
 
-    bool IsReverseV2Supported(const TensorInfo& input,
+    bool IsReverseV2Supported(const TensorInfo& input0,
+                              const TensorInfo& input1,
                               const TensorInfo& output,
-                              const ReverseV2Descriptor& descriptor,
                               Optional<std::string&> reasonIfUnsupported = EmptyOptional());
 
     bool IsShapeSupported(const TensorInfo& input,
diff --git a/include/armnn/Descriptors.hpp b/include/armnn/Descriptors.hpp
index 27ca50123f..9ff894f1b0 100644
--- a/include/armnn/Descriptors.hpp
+++ b/include/armnn/Descriptors.hpp
@@ -1620,28 +1620,4 @@ struct BatchMatMulDescriptor : BaseDescriptor
         const TensorShape& tensorShape);
 };
 
-struct ReverseV2Descriptor : BaseDescriptor
-{
-    ReverseV2Descriptor()
-        : m_Axis()
-        , m_MaxDimension(4)
-    {}
-
-    ReverseV2Descriptor(std::vector<int32_t> axis)
-        : m_Axis(axis)
-        , m_MaxDimension(4)
-    {}
-
-    bool operator ==(const ReverseV2Descriptor& rhs) const
-    {
-        return m_Axis == rhs.m_Axis;
-    }
-
-    /// The indices of the dimensions to reverse
-    std::vector<int32_t> m_Axis;
-    /// The max dimension supported in the lower levels of code
-    uint32_t m_MaxDimension;
-
-};
-
 } // namespace armnn
diff --git a/include/armnn/DescriptorsFwd.hpp b/include/armnn/DescriptorsFwd.hpp
index 4e9621d020..2c25a49f00 100644
--- a/include/armnn/DescriptorsFwd.hpp
+++ b/include/armnn/DescriptorsFwd.hpp
@@ -42,7 +42,6 @@ struct QLstmDescriptor;
 struct ReshapeDescriptor;
 struct ResizeDescriptor;
 struct ReduceDescriptor;
-struct ReverseV2Descriptor;
 struct SliceDescriptor;
 struct SoftmaxDescriptor;
 struct SpaceToBatchNdDescriptor;
diff --git a/include/armnn/INetwork.hpp b/include/armnn/INetwork.hpp
index e311fa8840..e20dd1c348 100644
--- a/include/armnn/INetwork.hpp
+++ b/include/armnn/INetwork.hpp
@@ -838,11 +838,9 @@ public:
                                            const char* name = nullptr);
 
     /// Add a ReverseV2 layer to the network
-    /// @param descriptor - Parameters for the ReverseV2 operation
     /// @param name - Optional name for the layer
     /// @return - Interface for configuring the layer
-    IConnectableLayer* AddReverseV2Layer(const ReverseV2Descriptor& descriptor,
-                                         const char* name = nullptr);
+    IConnectableLayer* AddReverseV2Layer(const char* name = nullptr);
 
     void ExecuteStrategy(IStrategy& strategy) const;
 
diff --git a/include/armnn/backends/WorkloadData.hpp b/include/armnn/backends/WorkloadData.hpp
index fe59fca795..e7d5e0e689 100644
--- a/include/armnn/backends/WorkloadData.hpp
+++ b/include/armnn/backends/WorkloadData.hpp
@@ -750,7 +750,7 @@ struct BatchMatMulQueueDescriptor : QueueDescriptorWithParameters<BatchMatMulDes
     void Validate(const WorkloadInfo& workloadInfo) const;
 };
 
-struct ReverseV2QueueDescriptor : QueueDescriptorWithParameters<ReverseV2Descriptor>
+struct ReverseV2QueueDescriptor : QueueDescriptor
 {
     void Validate(const WorkloadInfo& workloadInfo) const;
 };
diff --git a/src/armnn/BackendHelper.cpp b/src/armnn/BackendHelper.cpp
index 404d278efc..18184fbfb2 100644
--- a/src/armnn/BackendHelper.cpp
+++ b/src/armnn/BackendHelper.cpp
@@ -1211,16 +1211,16 @@ bool LayerSupportHandle::IsResizeSupported(const TensorInfo& input,
                                             reasonIfUnsupported);
 }
 
-bool LayerSupportHandle::IsReverseV2Supported(const armnn::TensorInfo &input,
+bool LayerSupportHandle::IsReverseV2Supported(const armnn::TensorInfo &input0,
+                                              const armnn::TensorInfo &input1,
                                               const armnn::TensorInfo &output,
-                                              const armnn::ReverseV2Descriptor &descriptor,
                                               Optional<std::string &> reasonIfUnsupported)
 {
-    TensorInfos infos{input, output};
+    TensorInfos infos{input0, input1, output};
 
     return m_LayerSupport->IsLayerSupported(LayerType::ReverseV2,
                                             infos,
-                                            descriptor,
+                                            BaseDescriptor(),
                                             EmptyOptional(),
                                             EmptyOptional(),
                                             reasonIfUnsupported);
diff --git a/src/armnn/Network.cpp b/src/armnn/Network.cpp
index ae5bde17ca..010fa0076b 100644
--- a/src/armnn/Network.cpp
+++ b/src/armnn/Network.cpp
@@ -639,10 +639,9 @@ IConnectableLayer* INetwork::AddBatchMatMulLayer(const BatchMatMulDescriptor &de
     return pNetworkImpl->AddBatchMatMulLayer(descriptor, name);
 }
 
-IConnectableLayer* INetwork::AddReverseV2Layer(const ReverseV2Descriptor &descriptor,
-                                               const char *name)
+IConnectableLayer* INetwork::AddReverseV2Layer(const char *name)
 {
-    return pNetworkImpl->AddReverseV2Layer(descriptor, name);
+    return pNetworkImpl->AddReverseV2Layer(name);
 }
 
 void INetwork::ExecuteStrategy(IStrategy& strategy) const
@@ -2930,9 +2929,9 @@ IConnectableLayer* NetworkImpl::AddBatchMatMulLayer(const BatchMatMulDescriptor&
     return m_Graph->AddLayer<BatchMatMulLayer>(desc, name);
 }
 
-IConnectableLayer* NetworkImpl::AddReverseV2Layer(const ReverseV2Descriptor &desc, const char *name)
+IConnectableLayer* NetworkImpl::AddReverseV2Layer(const char *name)
 {
-    return m_Graph->AddLayer<ReverseV2Layer>(desc, name);
+    return m_Graph->AddLayer<ReverseV2Layer>(name);
 }
 
 IConnectableLayer* NetworkImpl::AddPrecompiledLayer(const PreCompiledDescriptor& preCompiledDescriptor,
diff --git a/src/armnn/Network.hpp b/src/armnn/Network.hpp
index fc3ae42aa9..ae287f32d1 100644
--- a/src/armnn/Network.hpp
+++ b/src/armnn/Network.hpp
@@ -188,8 +188,7 @@ public:
     IConnectableLayer* AddReshapeLayer(const ReshapeDescriptor& reshapeDescriptor,
                                        const char* name = nullptr);
 
-    IConnectableLayer* AddReverseV2Layer(const ReverseV2Descriptor& ReverseV2Descriptor,
-                                         const char* name = nullptr);
+    IConnectableLayer* AddReverseV2Layer(const char* name = nullptr);
 
     IConnectableLayer* AddShapeLayer(const char* name = nullptr);
 
diff --git a/src/armnn/layers/ReverseV2Layer.cpp b/src/armnn/layers/ReverseV2Layer.cpp
index 201e19819b..e1160b6e16 100644
--- a/src/armnn/layers/ReverseV2Layer.cpp
+++ b/src/armnn/layers/ReverseV2Layer.cpp
@@ -10,9 +10,10 @@
 
 namespace armnn
 {
-ReverseV2Layer::ReverseV2Layer(const armnn::ReverseV2Descriptor &param, const char *name)
-    : LayerWithParameters(1, 1, LayerType::ReverseV2, param, name)
-{}
+ReverseV2Layer::ReverseV2Layer(const char* name)
+    : Layer(2, 1, LayerType::ReverseV2, name)
+{
+}
 
 std::unique_ptr<IWorkload> ReverseV2Layer::CreateWorkload(const armnn::IWorkloadFactory &factory) const
 {
@@ -24,27 +25,48 @@ std::unique_ptr<IWorkload> ReverseV2Layer::CreateWorkload(const armnn::IWorkload
 
 ReverseV2Layer* ReverseV2Layer::Clone(armnn::Graph &graph) const
 {
-    auto layer = CloneBase<ReverseV2Layer>(graph, m_Param, GetName());
+    auto layer = CloneBase<ReverseV2Layer>(graph, GetName());
 
     return std::move(layer);
 }
 
-/// Use the default Layer::InferOutputShape method
+std::vector<TensorShape> ReverseV2Layer::InferOutputShapes(const std::vector<TensorShape>& inputShapes) const
+{
+    ARMNN_ASSERT(inputShapes.size() == 2);
+
+    const auto inputDims = inputShapes[0].GetNumDimensions();
+
+    std::vector<unsigned int> dimSizes(inputDims);
+    for (unsigned i=0;  i<inputDims;  i++)
+    {
+        dimSizes[i] = inputShapes[0][i];
+    }
+
+    TensorShape outputShape({ inputDims, dimSizes.data() });
+
+    return std::vector<TensorShape>({ outputShape });
+}
 
 void ReverseV2Layer::ValidateTensorShapesFromInputs()
 {
-    VerifyLayerConnections(1, CHECK_LOCATION());
+    VerifyLayerConnections(2, CHECK_LOCATION());
 
     const TensorShape& outputShape = GetOutputSlot(0).GetTensorInfo().GetShape();
 
     VerifyShapeInferenceType(outputShape, m_ShapeInferenceMethod);
 
     auto inferredShapes = InferOutputShapes({
-        GetInputSlot(0).GetTensorInfo().GetShape() });
+        GetInputSlot(0).GetTensorInfo().GetShape(),
+        GetInputSlot(1).GetTensorInfo().GetShape()});
 
     ARMNN_ASSERT(inferredShapes.size() == 1);
 
     ValidateAndCopyShape(outputShape, inferredShapes[0], m_ShapeInferenceMethod, "ReverseV2Layer");
 }
 
-}
\ No newline at end of file
+void ReverseV2Layer::ExecuteStrategy(IStrategy& strategy) const
+{
+    strategy.ExecuteStrategy(this, BaseDescriptor(), {}, GetName());
+}
+
+}
diff --git a/src/armnn/layers/ReverseV2Layer.hpp b/src/armnn/layers/ReverseV2Layer.hpp
index 046670e9de..7dc0656aca 100644
--- a/src/armnn/layers/ReverseV2Layer.hpp
+++ b/src/armnn/layers/ReverseV2Layer.hpp
@@ -11,7 +11,7 @@ namespace armnn
 {
 
     /// This layer represents a ReverseV2 operation.
-    class ReverseV2Layer : public LayerWithParameters<ReverseV2Descriptor>
+    class ReverseV2Layer : public Layer
     {
     public:
         /// Makes a workload for the ReverseV2 type.
@@ -28,19 +28,18 @@ namespace armnn
         /// otherwise infers the output shapes from given input shapes and layer properties.
         /// @param [in] inputShapes The vector of input shapes for ReverseV2.
         /// @return A vector to the inferred output shape.
-
-        /// Use the default Layer::InferOutputShape method
-        // std::vector<TensorShape> InferOutputShapes(const std::vector<TensorShape>& inputShapes) const override;
+        std::vector<TensorShape> InferOutputShapes(const std::vector<TensorShape>& inputShapes) const override;
 
         /// Check if the input tensor shape(s)
         /// will lead to a valid configuration of @ref ReverseV2Layer.
         void ValidateTensorShapesFromInputs() override;
 
+        void ExecuteStrategy(IStrategy& strategy) const override;
+
     protected:
         /// Constructor to create a ReverseV2Layer.
-        /// @param [in] param ReverseV2Descriptor to configure the ReverseV2 operation.
         /// @param [in] name Optional name for the layer.
-        ReverseV2Layer(const ReverseV2Descriptor& param, const char* name);
+        ReverseV2Layer(const char* name);
 
         /// Default destructor
         ~ReverseV2Layer() = default;
diff --git a/src/armnnDeserializer/Deserializer.cpp b/src/armnnDeserializer/Deserializer.cpp
index d125ef56dd..1e40c637cc 100644
--- a/src/armnnDeserializer/Deserializer.cpp
+++ b/src/armnnDeserializer/Deserializer.cpp
@@ -2749,20 +2749,13 @@ void IDeserializer::DeserializerImpl::ParseReverseV2(GraphPtr graph, unsigned in
     CHECK_LAYERS(graph, 0, layerIndex);
 
     TensorRawPtrVector inputs = GetInputs(graph, layerIndex);
-    CHECK_VALID_SIZE(inputs.size(), 1);
+    CHECK_VALID_SIZE(inputs.size(), 2);
 
     TensorRawPtrVector outputs = GetOutputs(graph, layerIndex);
     CHECK_VALID_SIZE(outputs.size(), 1);
 
-    auto flatBufferDescriptor = graph->layers()->Get(layerIndex)->layer_as_ReverseV2Layer()->descriptor();
-    auto flatBufferAxis = flatBufferDescriptor->axis();
-
-    armnn::ReverseV2Descriptor descriptor;
-    descriptor.m_Axis =
-        std::vector<int32_t>(flatBufferAxis->begin(), flatBufferAxis->end());
-
     auto layerName = GetLayerName(graph, layerIndex);
-    IConnectableLayer* layer = m_Network->AddReverseV2Layer(descriptor, layerName.c_str());
+    IConnectableLayer* layer = m_Network->AddReverseV2Layer(layerName.c_str());
 
     armnn::TensorInfo outputTensorInfo = ToTensorInfo(outputs[0]);
     layer->GetOutputSlot(0).SetTensorInfo(outputTensorInfo);
diff --git a/src/armnnDeserializer/test/DeserializeReverseV2.cpp b/src/armnnDeserializer/test/DeserializeReverseV2.cpp
index 73816f8588..b6f207c9d1 100644
--- a/src/armnnDeserializer/test/DeserializeReverseV2.cpp
+++ b/src/armnnDeserializer/test/DeserializeReverseV2.cpp
@@ -14,61 +14,96 @@ TEST_SUITE("Deserializer_ReverseV2")
 {
     struct ReverseV2Fixture : public ParserFlatbuffersSerializeFixture
     {
-        explicit ReverseV2Fixture(const std::string& inputShape,
+        explicit ReverseV2Fixture(const std::string& inputShape0,
+                                  const std::string& inputShape1,
                                   const std::string& outputShape,
-                                  const std::string& dataType,
-                                  const std::string& axis)
+                                  const std::string& dataType0,
+                                  const std::string& dataType1)
         {
             m_JsonString = R"(
                 {
-                    inputIds: [0],
-                    outputIds: [2],
-                    layers: [
+                    inputIds: [0, 1],
+                    outputIds: [3],
+                    layers:
+                    [
                         {
                             layer_type: "InputLayer",
                             layer: {
-                                base: {
-                                    layerBindingId: 0,
-                                    base: {
-                                        index: 0,
-                                        layerName: "InputLayer",
-                                        layerType: "Input",
-                                        inputSlots: [{
-                                            index: 0,
-                                            connection: {sourceLayerIndex:0, outputSlotIndex:0 },
-                                        }],
-                                        outputSlots: [{
+                                  base: {
+                                        layerBindingId: 0,
+                                        base: {
                                             index: 0,
-                                            tensorInfo: {
-                                                dimensions: )" + inputShape + R"(,
-                                                dataType: )" + dataType + R"(
-                                            }
-                                        }]
-                                    }
-                                }
+                                            layerName: "InputLayer0",
+                                            layerType: "Input",
+                                            inputSlots:
+                                            [{
+                                                index: 0,
+                                                connection: {sourceLayerIndex:0, outputSlotIndex:0 },
+                                            }],
+                                            outputSlots:
+                                            [ {
+                                                index: 0,
+                                                tensorInfo: {
+                                                    dimensions: )" + inputShape0 + R"(,
+                                                    dataType: )" + dataType0 + R"(
+                                                }
+                                            }]
+                                        }
+                                  }
                             }
                         },
                         {
-                            layer_type: "ReverseV2Layer",
+                            layer_type: "InputLayer",
                             layer: {
-                                base: {
-                                    index: 1,
-                                    layerName: "ReverseV2Layer",
-                                    layerType: "ReverseV2",
-                                    inputSlots: [{
-                                        index: 0,
-                                        connection: {sourceLayerIndex:0, outputSlotIndex:0 },
-                                    }],
-                                    outputSlots: [{
-                                        index: 0,
-                                        tensorInfo: {
-                                            dimensions: )" + outputShape + R"(,
-                                            dataType: )" + dataType + R"(
+                                   base: {
+                                        layerBindingId: 1,
+                                        base: {
+                                              index:1,
+                                              layerName: "InputLayer1",
+                                              layerType: "Input",
+                                              inputSlots:
+                                              [{
+                                                  index: 0,
+                                                  connection: {sourceLayerIndex:0, outputSlotIndex:0 },
+                                              }],
+                                              outputSlots:
+                                              [{
+                                                  index: 0,
+                                                  tensorInfo: {
+                                                      dimensions: )" + inputShape1 + R"(,
+                                                      dataType: )" + dataType1 + R"(
+                                                  }
+                                              }]
                                         }
+                                   }
+                                }
+                        },
+                        {
+                            layer_type: "ReverseV2Layer",
+                            layer : {
+                                base: {
+                                     index:2,
+                                     layerName: "ReverseV2Layer",
+                                     layerType: "ReverseV2",
+                                     inputSlots:
+                                     [
+                                        {
+                                            index: 0,
+                                            connection: {sourceLayerIndex:0, outputSlotIndex:0 },
+                                         },
+                                         {
+                                            index: 1,
+                                            connection: {sourceLayerIndex:1, outputSlotIndex:0 },
+                                         }
+                                     ],
+                                     outputSlots:
+                                    [{
+                                         index: 0,
+                                         tensorInfo: {
+                                             dimensions: )" + outputShape + R"(,
+                                             dataType: )" + dataType0 + R"(
+                                         }
                                     }]
-                                },
-                                descriptor: {
-                                    axis: )" + axis + R"(
                                 }
                             }
                         },
@@ -76,31 +111,32 @@ TEST_SUITE("Deserializer_ReverseV2")
                             layer_type: "OutputLayer",
                             layer: {
                                 base:{
-                                    layerBindingId: 2,
-                                    base: {
-                                        index: 2,
-                                        layerName: "OutputLayer",
-                                        layerType: "Output",
-                                        inputSlots: [{
-                                            index: 0,
-                                            connection: {sourceLayerIndex:1, outputSlotIndex:0 },
-                                        }],
-                                        outputSlots: [{
-                                            index: 0,
-                                            tensorInfo: {
-                                                dimensions: )" + outputShape + R"(,
-                                                dataType: )" + dataType + R"(
-                                            },
-                                        }],
-                                    }
+                                      layerBindingId: 0,
+                                      base: {
+                                            index: 3,
+                                            layerName: "OutputLayer",
+                                            layerType: "Output",
+                                            inputSlots:
+                                            [{
+                                                index: 0,
+                                                connection: {sourceLayerIndex:2, outputSlotIndex:0 },
+                                            }],
+                                            outputSlots:
+                                            [{
+                                                index: 0,
+                                                tensorInfo: {
+                                                    dimensions: )" + outputShape + R"(,
+                                                    dataType: )" + dataType0 + R"(
+                                                }
+                                            }]
+                                      }
                                 }
                             }
                         }
                     ]
-                }
-            )";
+                } )";
 
-            SetupSingleInputSingleOutput("InputLayer", "OutputLayer");
+            Setup();
         }
     };
 
@@ -109,65 +145,155 @@ TEST_SUITE("Deserializer_ReverseV2")
     struct SimpleReverseV2FixtureFloat32 : ReverseV2Fixture
     {
         SimpleReverseV2FixtureFloat32()
+                : ReverseV2Fixture("[ 2, 2 ]",
+                                   "[ 1 ]",
+                                   "[ 2, 2 ]",
+                                   "Float32",
+                                   "Signed32")
+        {}
+    };
+
+    TEST_CASE_FIXTURE(SimpleReverseV2FixtureFloat32, "SimpleReverseV2TestFloat32")
+    {
+        RunTest<2, armnn::DataType::Float32, armnn::DataType::Signed32, armnn::DataType::Float32>(
+            0,
+            {
+                {
+                    "InputLayer0",
+                    { 1.0f, 2.0f,
+                      3.0f, 4.0f }
+                }
+            },
+            {
+                {
+                    "InputLayer1",
+                    { 1 }
+                }
+            },
+            {
+                {
+                    "OutputLayer",
+                    { 2.0f, 1.0f,
+                      4.0f, 3.0f }
+                }
+            }
+        );
+    }
+
+    struct SimpleReverseV2FixtureFloat32OtherAxis : ReverseV2Fixture
+    {
+        SimpleReverseV2FixtureFloat32OtherAxis()
             : ReverseV2Fixture("[ 2, 2 ]",
+                               "[ 1 ]",
                                "[ 2, 2 ]",
                                "Float32",
-                               "[1]")
+                               "Signed32")
         {}
     };
 
-    TEST_CASE_FIXTURE(SimpleReverseV2FixtureFloat32, "SimpleReverseV2TestFloat32")
+    TEST_CASE_FIXTURE(SimpleReverseV2FixtureFloat32OtherAxis, "SimpleReverseV2FixtureFloat32OtherAxis")
     {
-        RunTest<4, armnn::DataType::Float32>(
+        RunTest<2, armnn::DataType::Float32, armnn::DataType::Signed32, armnn::DataType::Float32>(
             0,
-            { 1.0f, 2.0f,
-              3.0f, 4.0f },
-            { 2.0f, 1.0f,
-              4.0f, 3.0f }
-            );
+            {
+                {
+                    "InputLayer0",
+                    { 1.0f, 2.0f,
+                      3.0f, 4.0f }
+                }
+            },
+            {
+                {
+                    "InputLayer1",
+                    { 1 }
+                }
+            },
+            {
+                {
+                    "OutputLayer",
+                    { 2.0f, 1.0f,
+                      4.0f, 3.0f }
+                }
+            }
+        );
     }
 
-    struct SimpleReverseV2FixtureFloat32ZeroAxis : ReverseV2Fixture
+    struct SimpleReverseV2FixtureFloat32NegativeFirstAxis : ReverseV2Fixture
     {
-        SimpleReverseV2FixtureFloat32ZeroAxis()
+        SimpleReverseV2FixtureFloat32NegativeFirstAxis()
                 : ReverseV2Fixture("[ 2, 2 ]",
+                                   "[ 1 ]",
                                    "[ 2, 2 ]",
                                    "Float32",
-                                   "[0]")
+                                   "Signed32")
         {}
     };
 
-    TEST_CASE_FIXTURE(SimpleReverseV2FixtureFloat32ZeroAxis, "SimpleReverseV2TestFloat32ZeroAxis")
+    TEST_CASE_FIXTURE(SimpleReverseV2FixtureFloat32NegativeFirstAxis, "SimpleReverseV2FixtureFloat32NegativeFirstAxis")
     {
-        RunTest<4, armnn::DataType::Float32>(
-                0,
-                { 1.0f, 2.0f,
-                  3.0f, 4.0f },
-                { 3.0f, 4.0f,
-                  1.0f, 2.0f }
+        RunTest<2, armnn::DataType::Float32, armnn::DataType::Signed32, armnn::DataType::Float32>(
+            0,
+            {
+                {
+                    "InputLayer0",
+                    { 1.0f, 2.0f,
+                      3.0f, 4.0f }
+                }
+            },
+            {
+                {
+                    "InputLayer1",
+                    { -2 }
+                }
+            },
+            {
+                {
+                    "OutputLayer",
+                    { 3.0f, 4.0f,
+                      1.0f, 2.0f }
+                }
+            }
         );
     }
 
-    struct SimpleReverseV2FixtureFloat32NegativeAxis : ReverseV2Fixture
+    struct SimpleReverseV2FixtureFloat32NegativeSecondAxis : ReverseV2Fixture
     {
-        SimpleReverseV2FixtureFloat32NegativeAxis()
+        SimpleReverseV2FixtureFloat32NegativeSecondAxis()
                 : ReverseV2Fixture("[ 3, 3 ]",
+                                   "[ 1 ]",
                                    "[ 3, 3 ]",
                                    "Float32",
-                                   "[-1]")
+                                   "Signed32")
         {}
     };
 
-    TEST_CASE_FIXTURE(SimpleReverseV2FixtureFloat32NegativeAxis, "SimpleReverseV2TestFloat32NegativeAxis")
+    TEST_CASE_FIXTURE(SimpleReverseV2FixtureFloat32NegativeSecondAxis,
+        "SimpleReverseV2FixtureFloat32NegativeSecondAxis")
     {
-        RunTest<4, armnn::DataType::Float32>(
-                0,
-                { 1.0f, 2.0f, 3.0f,
-                  4.0f, 5.0f, 6.0f,
-                  7.0f, 8.0f, 9.0f },
-                { 3.0f, 2.0f, 1.0f,
-                  6.0f, 5.0f, 4.0f,
-                  9.0f, 8.0f, 7.0f }
+        RunTest<2, armnn::DataType::Float32, armnn::DataType::Signed32, armnn::DataType::Float32>(
+            0,
+            {
+                {
+                    "InputLayer0",
+                    { 1.0f, 2.0f, 3.0f,
+                      4.0f, 5.0f, 6.0f,
+                      7.0f, 8.0f, 9.0f }
+                }
+            },
+            {
+                {
+                    "InputLayer1",
+                    { -1 }
+                }
+            },
+            {
+                {
+                    "OutputLayer",
+                    { 3.0f, 2.0f, 1.0f,
+                      6.0f, 5.0f, 4.0f,
+                      9.0f, 8.0f, 7.0f }
+                }
+            }
         );
     }
 
@@ -175,38 +301,55 @@ TEST_SUITE("Deserializer_ReverseV2")
     {
         SimpleReverseV2FixtureFloat32ThreeAxis()
                 : ReverseV2Fixture("[ 3, 3, 3 ]",
+                                   "[ 3 ]",
                                    "[ 3, 3, 3 ]",
                                    "Float32",
-                                   "[0, 2, 1]")
+                                   "Signed32")
         {}
     };
 
     TEST_CASE_FIXTURE(SimpleReverseV2FixtureFloat32ThreeAxis, "SimpleReverseV2TestFloat32ThreeAxis")
     {
-        RunTest<4, armnn::DataType::Float32>(
-                0,
-                { 1.0f, 2.0f, 3.0f,
-                  4.0f, 5.0f, 6.0f,
-                  7.0f, 8.0f, 9.0f,
-
-                  11.0f, 12.0f, 13.0f,
-                  14.0f, 15.0f, 16.0f,
-                  17.0f, 18.0f, 19.0f,
-
-                  21.0f, 22.0f, 23.0f,
-                  24.0f, 25.0f, 26.0f,
-                  27.0f, 28.0f, 29.0f },
-                { 29.0f, 28.0f, 27.0f,
-                  26.0f, 25.0f, 24.0f,
-                  23.0f, 22.0f, 21.0f,
-
-                  19.0f, 18.0f, 17.0f,
-                  16.0f, 15.0f, 14.0f,
-                  13.0f, 12.0f, 11.0f,
-
-                  9.0f, 8.0f, 7.0f,
-                  6.0f, 5.0f, 4.0f,
-                  3.0f, 2.0f, 1.0f }
+        RunTest<2, armnn::DataType::Float32, armnn::DataType::Signed32, armnn::DataType::Float32>(
+            0,
+            {
+                {
+                    "InputLayer0",
+                    { 1.0f, 2.0f, 3.0f,
+                      4.0f, 5.0f, 6.0f,
+                      7.0f, 8.0f, 9.0f,
+
+                      11.0f, 12.0f, 13.0f,
+                      14.0f, 15.0f, 16.0f,
+                      17.0f, 18.0f, 19.0f,
+
+                      21.0f, 22.0f, 23.0f,
+                      24.0f, 25.0f, 26.0f,
+                      27.0f, 28.0f, 29.0f },
+                }
+            },
+            {
+                {
+                    "InputLayer1",
+                    { 0, 2, 1 }
+                }
+            },
+            {
+                {
+                    "OutputLayer",
+                    { 29.0f, 28.0f, 27.0f,
+                      26.0f, 25.0f, 24.0f,
+                      23.0f, 22.0f, 21.0f,
+
+                      19.0f, 18.0f, 17.0f,
+                      16.0f, 15.0f, 14.0f,
+                      13.0f, 12.0f, 11.0f,
+
+                      9.0f, 8.0f, 7.0f,
+                      6.0f, 5.0f, 4.0f,
+                      3.0f, 2.0f, 1.0f }
+                }
+            }
         );
     }
 
@@ -214,38 +357,55 @@ TEST_SUITE("Deserializer_ReverseV2")
     {
         SimpleReverseV2FixtureQuantisedAsymm8ThreeAxis()
                 : ReverseV2Fixture("[ 3, 3, 3 ]",
+                                   "[ 3 ]",
                                    "[ 3, 3, 3 ]",
                                    "QuantisedAsymm8",
-                                   "[0, 2, 1]")
+                                   "Signed32")
         {}
     };
 
     TEST_CASE_FIXTURE(SimpleReverseV2FixtureQuantisedAsymm8ThreeAxis, "SimpleReverseV2TestQuantisedAsymm8ThreeAxis")
     {
-        RunTest<4, armnn::DataType::QAsymmU8>(
-                0,
-                { 1, 2, 3,
-                  4, 5, 6,
-                  7, 8, 9,
-
-                  11, 12, 13,
-                  14, 15, 16,
-                  17, 18, 19,
-
-                  21, 22, 23,
-                  24, 25, 26,
-                  27, 28, 29 },
-                { 29, 28, 27,
-                  26, 25, 24,
-                  23, 22, 21,
-
-                  19, 18, 17,
-                  16, 15, 14,
-                  13, 12, 11,
-
-                  9, 8, 7,
-                  6, 5, 4,
-                  3, 2, 1 }
+        RunTest<2, armnn::DataType::QAsymmU8, armnn::DataType::Signed32, armnn::DataType::QAsymmU8>(
+            0,
+            {
+                {
+                    "InputLayer0",
+                    { 1, 2, 3,
+                      4, 5, 6,
+                      7, 8, 9,
+
+                      11, 12, 13,
+                      14, 15, 16,
+                      17, 18, 19,
+
+                      21, 22, 23,
+                      24, 25, 26,
+                      27, 28, 29 },
+                }
+            },
+            {
+                {
+                    "InputLayer1",
+                    { 0, 2, 1 }
+                }
+            },
+            {
+                {
+                    "OutputLayer",
+                    { 29, 28, 27,
+                      26, 25, 24,
+                      23, 22, 21,
+
+                      19, 18, 17,
+                      16, 15, 14,
+                      13, 12, 11,
+
+                      9, 8, 7,
+                      6, 5, 4,
+                      3, 2, 1 }
+                }
+            }
         );
     }
 }
diff --git a/src/armnnDeserializer/test/ParserFlatbuffersSerializeFixture.hpp b/src/armnnDeserializer/test/ParserFlatbuffersSerializeFixture.hpp
index 31ff026887..0b717bc0fd 100644
--- a/src/armnnDeserializer/test/ParserFlatbuffersSerializeFixture.hpp
+++ b/src/armnnDeserializer/test/ParserFlatbuffersSerializeFixture.hpp
@@ -1,5 +1,5 @@
 //
-// Copyright © 2017 Arm Ltd. All rights reserved.
+// Copyright © 2019-2023 Arm Ltd. All rights reserved.
 // SPDX-License-Identifier: MIT
 //
 
@@ -150,6 +150,18 @@ struct ParserFlatbuffersSerializeFixture
                  const std::map<std::string, std::vector<InputDataType>>& inputData,
                  const std::map<std::string, std::vector<OutputDataType>>& expectedOutputData);
 
+    template<std::size_t NumOutputDimensions,
+            armnn::DataType ArmnnInputType0,
+            armnn::DataType ArmnnInputType1,
+            armnn::DataType ArmnnOutputType,
+            typename InputDataType0 = armnn::ResolveType<ArmnnInputType0>,
+            typename InputDataType1 = armnn::ResolveType<ArmnnInputType1>,
+            typename OutputDataType = armnn::ResolveType<ArmnnOutputType>>
+    void RunTest(unsigned int layersId,
+                 const std::map<std::string, std::vector<InputDataType0>>& inputData0,
+                 const std::map<std::string, std::vector<InputDataType1>>& inputData1,
+                 const std::map<std::string, std::vector<OutputDataType>>& expectedOutputData);
+
     void CheckTensors(const TensorRawPtr& tensors, size_t shapeSize, const std::vector<int32_t>& shape,
                       armnnSerializer::TensorInfo tensorType, const std::string& name,
                       const float scale, const int64_t zeroPoint)
@@ -246,3 +258,68 @@ void ParserFlatbuffersSerializeFixture::RunTest(
         CHECK_MESSAGE(result.m_Result, result.m_Message.str());
     }
 }
+
+template<std::size_t NumOutputDimensions,
+        armnn::DataType ArmnnInputType0,
+        armnn::DataType ArmnnInputType1,
+        armnn::DataType ArmnnOutputType,
+        typename InputDataType0,
+        typename InputDataType1,
+        typename OutputDataType>
+void ParserFlatbuffersSerializeFixture::RunTest(
+        unsigned int layersId,
+        const std::map<std::string, std::vector<InputDataType0>>& inputData0,
+        const std::map<std::string, std::vector<InputDataType1>>& inputData1,
+        const std::map<std::string, std::vector<OutputDataType>>& expectedOutputData)
+{
+    auto ConvertBindingInfo = [](const armnnDeserializer::BindingPointInfo& bindingInfo)
+    {
+        return std::make_pair(bindingInfo.m_BindingId, bindingInfo.m_TensorInfo);
+    };
+
+    // Setup the armnn input tensors from the given vectors.
+    armnn::InputTensors inputTensors;
+    for (auto&& it : inputData0)
+    {
+        armnn::BindingPointInfo bindingInfo = ConvertBindingInfo(
+            m_Parser->GetNetworkInputBindingInfo(layersId, it.first));
+        bindingInfo.second.SetConstant(true);
+        armnn::VerifyTensorInfoDataType(bindingInfo.second, ArmnnInputType0);
+        inputTensors.push_back({ bindingInfo.first, armnn::ConstTensor(bindingInfo.second, it.second.data()) });
+    }
+
+    for (auto&& it : inputData1)
+    {
+        armnn::BindingPointInfo bindingInfo = ConvertBindingInfo(
+            m_Parser->GetNetworkInputBindingInfo(layersId, it.first));
+        bindingInfo.second.SetConstant(true);
+        armnn::VerifyTensorInfoDataType(bindingInfo.second, ArmnnInputType1);
+        inputTensors.push_back({ bindingInfo.first, armnn::ConstTensor(bindingInfo.second, it.second.data()) });
+    }
+
+    // Allocate storage for the output tensors to be written to and setup the armnn output tensors.
+    std::map<std::string, std::vector<OutputDataType>> outputStorage;
+    armnn::OutputTensors outputTensors;
+    for (auto&& it : expectedOutputData)
+    {
+        armnn::BindingPointInfo bindingInfo = ConvertBindingInfo(
+            m_Parser->GetNetworkOutputBindingInfo(layersId, it.first));
+        armnn::VerifyTensorInfoDataType(bindingInfo.second, ArmnnOutputType);
+        outputStorage.emplace(it.first, std::vector<OutputDataType>(bindingInfo.second.GetNumElements()));
+        outputTensors.push_back(
+            { bindingInfo.first, armnn::Tensor(bindingInfo.second, outputStorage.at(it.first).data()) });
+    }
+
+    m_Runtime->EnqueueWorkload(m_NetworkIdentifier, inputTensors, outputTensors);
+
+    // Compare each output tensor to the expected values
+    for (auto&& it : expectedOutputData)
+    {
+        armnn::BindingPointInfo bindingInfo = ConvertBindingInfo(
+            m_Parser->GetNetworkOutputBindingInfo(layersId, it.first));
+        auto outputExpected = it.second;
+        auto result = CompareTensors(outputExpected, outputStorage[it.first],
+                                     bindingInfo.second.GetShape(), bindingInfo.second.GetShape());
+        CHECK_MESSAGE(result.m_Result, result.m_Message.str());
+    }
+}
diff --git a/src/armnnSerializer/ArmnnSchema.fbs b/src/armnnSerializer/ArmnnSchema.fbs
index 995a6013c1..75dd252e68 100644
--- a/src/armnnSerializer/ArmnnSchema.fbs
+++ b/src/armnnSerializer/ArmnnSchema.fbs
@@ -984,11 +984,6 @@ table ResizeDescriptor {
 
 table ReverseV2Layer {
     base:LayerBase;
-    descriptor:ReverseV2Descriptor;
-}
-
-table ReverseV2Descriptor {
-    axis:[int];
 }
 
 table StackLayer {
diff --git a/src/armnnSerializer/Serializer.cpp b/src/armnnSerializer/Serializer.cpp
index e10b66f51d..39a42954a9 100644
--- a/src/armnnSerializer/Serializer.cpp
+++ b/src/armnnSerializer/Serializer.cpp
@@ -1036,20 +1036,14 @@ void SerializerStrategy::SerializeResizeLayer(const armnn::IConnectableLayer* la
 }
 
 void SerializerStrategy::SerializeReverseV2Layer(const armnn::IConnectableLayer* layer,
-                                              const armnn::ReverseV2Descriptor& reverseV2Descriptor,
-                                              const char* name)
+                                                 const char* name)
 {
     IgnoreUnused(name);
 
     auto flatBufferBaseLayer = CreateLayerBase(layer, serializer::LayerType::LayerType_ReverseV2);
 
-    auto flatBufferDescriptor =
-        CreateReverseV2Descriptor(m_flatBufferBuilder,
-                                  m_flatBufferBuilder.CreateVector(reverseV2Descriptor.m_Axis));
-
     auto flatBufferLayer = serializer::CreateReverseV2Layer(m_flatBufferBuilder,
-                                                            flatBufferBaseLayer,
-                                                            flatBufferDescriptor);
+                                                            flatBufferBaseLayer);
 
     CreateAnyLayer(flatBufferLayer.o, serializer::Layer::Layer_ReverseV2Layer);
 }
@@ -2357,9 +2351,7 @@ void SerializerStrategy::ExecuteStrategy(const armnn::IConnectableLayer* layer,
         }
         case armnn::LayerType::ReverseV2:
         {
-            const armnn::ReverseV2Descriptor& layerDescriptor =
-                    static_cast<const armnn::ReverseV2Descriptor&>(descriptor);
-            SerializeReverseV2Layer(layer, layerDescriptor, name);
+            SerializeReverseV2Layer(layer, name);
             break;
         }
         case armnn::LayerType::Shape:
diff --git a/src/armnnSerializer/Serializer.hpp b/src/armnnSerializer/Serializer.hpp
index eb724752f2..9e9eca8e26 100644
--- a/src/armnnSerializer/Serializer.hpp
+++ b/src/armnnSerializer/Serializer.hpp
@@ -290,7 +290,6 @@ private:
                               const char* name = nullptr);
 
     void SerializeReverseV2Layer(const armnn::IConnectableLayer* layer,
-                                 const armnn::ReverseV2Descriptor& reverseV2Descriptor,
                                  const char* name = nullptr);
 
     void SerializeSliceLayer(const armnn::IConnectableLayer* layer,
diff --git a/src/armnnSerializer/test/SerializerTests.cpp b/src/armnnSerializer/test/SerializerTests.cpp
index 49971d2ecc..163e5c8374 100644
--- a/src/armnnSerializer/test/SerializerTests.cpp
+++ b/src/armnnSerializer/test/SerializerTests.cpp
@@ -2457,26 +2457,27 @@ TEST_CASE("SerializeReverseV2")
 {
     const std::string layerName("reverseV2");
     const armnn::TensorInfo inputInfo  = armnn::TensorInfo({2, 3, 4}, armnn::DataType::Float32);
+    const armnn::TensorInfo axisInfo   = armnn::TensorInfo({3}, armnn::DataType::Signed32, 0.0f, 0, true);
     const armnn::TensorInfo outputInfo = armnn::TensorInfo({2, 3, 4}, armnn::DataType::Float32);
 
-    armnn::ReverseV2Descriptor desc;
-    desc.m_Axis = {1, 0, 2};
-
     armnn::INetworkPtr network = armnn::INetwork::Create();
     armnn::IConnectableLayer* const inputLayer = network->AddInputLayer(0);
-    armnn::IConnectableLayer* const reverseV2Layer = network->AddReverseV2Layer(desc, layerName.c_str());
+    armnn::IConnectableLayer* const axisLayer = network->AddInputLayer(1);
+    armnn::IConnectableLayer* const reverseV2Layer = network->AddReverseV2Layer(layerName.c_str());
     armnn::IConnectableLayer* const outputLayer = network->AddOutputLayer(0);
 
     inputLayer->GetOutputSlot(0).Connect(reverseV2Layer->GetInputSlot(0));
+    axisLayer->GetOutputSlot(0).Connect(reverseV2Layer->GetInputSlot(1));
     reverseV2Layer->GetOutputSlot(0).Connect(outputLayer->GetInputSlot(0));
 
     inputLayer->GetOutputSlot(0).SetTensorInfo(inputInfo);
+    axisLayer->GetOutputSlot(0).SetTensorInfo(axisInfo);
     reverseV2Layer->GetOutputSlot(0).SetTensorInfo(outputInfo);
 
     armnn::INetworkPtr deserializedNetwork = DeserializeNetwork(SerializeNetwork(*network));
     CHECK(deserializedNetwork);
 
-    LayerVerifierBaseWithDescriptor<armnn::ReverseV2Descriptor> verifier(layerName, {inputInfo}, {outputInfo}, desc);
+    LayerVerifierBase verifier(layerName, {inputInfo, axisInfo}, {outputInfo});
     deserializedNetwork->ExecuteStrategy(verifier);
 }
 
diff --git a/src/armnnTfLiteParser/TfLiteParser.cpp b/src/armnnTfLiteParser/TfLiteParser.cpp
index 078786cb93..77ce565959 100644
--- a/src/armnnTfLiteParser/TfLiteParser.cpp
+++ b/src/armnnTfLiteParser/TfLiteParser.cpp
@@ -3294,46 +3294,13 @@ void TfLiteParserImpl::ParseReverseV2(size_t subgraphIndex, size_t operatorIndex
     TensorInfo axisTensorInfo = ToTensorInfo(inputs[1]);
     TensorInfo outputTensorInfo = ToTensorInfo(outputs[0]);
 
-    std::vector<int32_t> axisTensorData(axisTensorInfo.GetNumElements());
-
-    BufferRawPtr axisBufferPtr = GetBuffer(m_Model, inputs[1]->buffer);
-    ::memcpy(axisTensorData.data(), axisBufferPtr->data.data(), axisTensorInfo.GetNumBytes());
-
-    ReverseV2Descriptor descriptor(axisTensorData);
-
-    auto inputRank = static_cast<int32_t>(inputTensorInfo.GetNumDimensions());
-    std::vector<bool> dimFlag(inputRank, false);
-
-    for (auto axis : axisTensorData)
-    {
-        if (axis < -inputRank || axis >= inputRank)
-        {
-            throw ParseException(
-                fmt::format("Operation has invalid axis: {} It is out of bounds [ -{}, {} ) {}",
-                            axis,
-                            inputRank, inputRank,
-                            CHECK_LOCATION().AsString()));
-        }
-
-        auto posAxis = axis < 0 ? axis + inputRank : axis;
-
-        if (dimFlag[posAxis])
-        {
-            throw ParseException(
-                fmt::format("Operation has repeated axis: {} {}",
-                            axis,
-                            CHECK_LOCATION().AsString()));
-        }
-        dimFlag[posAxis] = true;
-    }
-
-    IConnectableLayer* layer = m_Network->AddReverseV2Layer(descriptor, layerName.c_str());
+    IConnectableLayer* layer = m_Network->AddReverseV2Layer(layerName.c_str());
     ARMNN_ASSERT(layer != nullptr);
 
     layer->GetOutputSlot(0).SetTensorInfo(outputTensorInfo);
 
     auto inputTensorIndexes = AsUnsignedVector(GetInputTensorIds(m_Model, subgraphIndex, operatorIndex));
-    RegisterInputSlots(subgraphIndex, operatorIndex, layer, {inputTensorIndexes[0]});
+    RegisterInputSlots(subgraphIndex, operatorIndex, layer, {inputTensorIndexes[0], inputTensorIndexes[1]});
 
     auto outputTensorIndexes = AsUnsignedVector(GetOutputTensorIds(m_Model, subgraphIndex, operatorIndex));
     RegisterOutputSlots(subgraphIndex, operatorIndex, layer, {outputTensorIndexes[0]});
diff --git a/src/armnnTfLiteParser/test/ReverseV2.cpp b/src/armnnTfLiteParser/test/ReverseV2.cpp
index 9604a3a1de..7dad19a40d 100644
--- a/src/armnnTfLiteParser/test/ReverseV2.cpp
+++ b/src/armnnTfLiteParser/test/ReverseV2.cpp
@@ -12,7 +12,6 @@ struct ReverseV2Fixture : public ParserFlatbuffersFixture
     explicit ReverseV2Fixture(const std::string& inputShape,
                               const std::string& outputShape,
                               const std::string& axisShape,
-                              const std::string& axisData,
                               const std::string& dataType = "FLOAT32",
                               const std::string& scale = "1.0",
                               const std::string& offset = "0")
@@ -75,8 +74,8 @@ struct ReverseV2Fixture : public ParserFlatbuffersFixture
                 } ],
                 "buffers" : [
                     { },
-                    { "data": )" + axisData + R"(, },
                     { },
+                    { }
                 ]
             }
         )";
@@ -86,15 +85,30 @@ struct ReverseV2Fixture : public ParserFlatbuffersFixture
 
 struct SimpleReverseV2Fixture : public ReverseV2Fixture
 {
-    SimpleReverseV2Fixture() : ReverseV2Fixture("[ 2, 2, 2 ]", "[ 2, 2, 2 ]", "[ 2 ]", "[ 0,0,0,0, 1,0,0,0 ]") {}
+    SimpleReverseV2Fixture() : ReverseV2Fixture("[ 2, 2, 2 ]", "[ 2, 2, 2 ]", "[ 2 ]" ) {}
 };
 
 TEST_CASE_FIXTURE(SimpleReverseV2Fixture, "ParseReverseV2")
 {
-    RunTest<3, armnn::DataType::Float32>
+    RunTest<3, armnn::DataType::Float32, armnn::DataType::Signed32, armnn::DataType::Float32>
         (0,
         {{ "inputTensor",  { 1, 2, 3, 4, 5, 6, 7, 8 }}},
+        {{ "axis",  { 0, 1 }}},
         {{ "outputTensor", { 7, 8, 5, 6, 3, 4, 1, 2 }}});
 }
 
-}
\ No newline at end of file
+struct SimpleReverseV2FixtureNegativeAxis : public ReverseV2Fixture
+{
+    SimpleReverseV2FixtureNegativeAxis() : ReverseV2Fixture("[ 2, 2, 2 ]", "[ 2, 2, 2 ]", "[ 1 ]" ) {}
+};
+
+TEST_CASE_FIXTURE(SimpleReverseV2FixtureNegativeAxis, "ParseReverseV2")
+{
+    RunTest<3, armnn::DataType::Float32, armnn::DataType::Signed32, armnn::DataType::Float32>
+        (0,
+        {{ "inputTensor",  { 1, 2, 3, 4, 5, 6, 7, 8 }}},
+        {{ "axis",  { -1 }}},
+        {{ "outputTensor", { 2, 1, 4, 3, 6, 5, 8, 7  }}});
+}
+
+}
diff --git a/src/backends/backendsCommon/WorkloadData.cpp b/src/backends/backendsCommon/WorkloadData.cpp
index a26aaf490b..bd3c7c2760 100644
--- a/src/backends/backendsCommon/WorkloadData.cpp
+++ b/src/backends/backendsCommon/WorkloadData.cpp
@@ -1617,18 +1617,35 @@ void ResizeQueueDescriptor::Validate(const WorkloadInfo& workloadInfo) const
 void ReverseV2QueueDescriptor::Validate(const WorkloadInfo &workloadInfo) const {
     const std::string descriptorName{"ReverseV2QueueDescriptor"};
 
-    ValidateNumInputs(workloadInfo,  descriptorName, 1);
+    // Backend restriction
+    const unsigned int maxDimensions = 4;
+
+    ValidateNumInputs(workloadInfo,  descriptorName, 2);
     ValidateNumOutputs(workloadInfo, descriptorName, 1);
 
     const TensorInfo& inputTensorInfo  = workloadInfo.m_InputTensorInfos[0];
+    const TensorInfo& axisTensorInfo  = workloadInfo.m_InputTensorInfos[1];
     const TensorInfo& outputTensorInfo = workloadInfo.m_OutputTensorInfos[0];
 
-    auto inputTensorNumDimensions = inputTensorInfo.GetNumDimensions();
-    if (inputTensorNumDimensions > m_Parameters.m_MaxDimension)
+    const auto inputTensorNumDimensions = inputTensorInfo.GetNumDimensions();
+    if (inputTensorNumDimensions > maxDimensions)
     {
         throw InvalidArgumentException(descriptorName +
             ": Input tensors with rank greater than " +
-            std::to_string(m_Parameters.m_MaxDimension) + " are not supported.");
+            std::to_string(maxDimensions) + " are not supported.");
+    }
+
+    const auto axisTensorNumDimensions = axisTensorInfo.GetNumDimensions();
+    if (axisTensorNumDimensions > maxDimensions)
+    {
+        throw InvalidArgumentException(descriptorName +
+            ": More than " + std::to_string(maxDimensions) + " axes cannot be specified.");
+    }
+
+    if (axisTensorNumDimensions > inputTensorNumDimensions)
+    {
+        throw InvalidArgumentException(descriptorName +
+            ": More axes specified than the number of axes on the input tensor.");
     }
 
     std::vector<DataType> supportedTypes =
@@ -1642,44 +1659,18 @@ void ReverseV2QueueDescriptor::Validate(const WorkloadInfo &workloadInfo) const
     };
 
     ValidateDataTypes(inputTensorInfo, supportedTypes, descriptorName);
-    ValidateTensorShapesMatch(inputTensorInfo, outputTensorInfo, descriptorName, "input", "output");
-    ValidateTensorDataTypesMatch(inputTensorInfo, outputTensorInfo, descriptorName, "input", "output");
 
-    if (m_Parameters.m_Axis.size() > inputTensorNumDimensions)
-    {
-        throw InvalidArgumentException(descriptorName + ": More axes specified than is on the input tensor.");
-    }
-    if (m_Parameters.m_Axis.size() > m_Parameters.m_MaxDimension)
+    std::vector<DataType> axisSupportedTypes =
     {
-        throw InvalidArgumentException(descriptorName +
-            ": More than " + std::to_string(m_Parameters.m_MaxDimension) + " axes cannot be specified.");
-    }
+        DataType::Signed32,
+    };
 
-    if (! m_Parameters.m_Axis.empty())
-    {
-        // First check that we have unique axis values
-        auto checkAxis = m_Parameters.m_Axis;
-        std::sort(checkAxis.begin(), checkAxis.end());
-        auto lastUnique = std::unique(checkAxis.begin(), checkAxis.end());
-        if (lastUnique != checkAxis.end())
-        {
-            throw InvalidArgumentException(descriptorName + ": Axes values must be unique.");
-        }
+    ValidateDataTypes(axisTensorInfo, axisSupportedTypes, descriptorName);
 
-        // Next check that the axes values are in range: [-rank, rank]
-        const auto minmax =
-                std::minmax_element(std::begin(m_Parameters.m_Axis), std::end(m_Parameters.m_Axis));
-        if (((*minmax.first) < int32_t(-inputTensorNumDimensions)) ||
-            ((*minmax.second) >= int32_t (inputTensorNumDimensions)))
-        {
-            throw InvalidArgumentException(descriptorName +
-                ": Axes values must in range [-" + std::to_string(inputTensorNumDimensions) + "," +
-                std::to_string(inputTensorNumDimensions) + "].");
-        }
-    }
+    ValidateTensorShapesMatch(inputTensorInfo, outputTensorInfo, descriptorName, "input", "output");
+    ValidateTensorDataTypesMatch(inputTensorInfo, outputTensorInfo, descriptorName, "input", "output");
 }
 
-
 void FakeQuantizationQueueDescriptor::Validate(const WorkloadInfo& workloadInfo) const
 {
     const std::string descriptorName{"FakeQuantizationQueueDescriptor"};
diff --git a/src/backends/backendsCommon/WorkloadFactory.cpp b/src/backends/backendsCommon/WorkloadFactory.cpp
index ac4bcc90f6..ee797b632c 100644
--- a/src/backends/backendsCommon/WorkloadFactory.cpp
+++ b/src/backends/backendsCommon/WorkloadFactory.cpp
@@ -1121,12 +1121,12 @@ bool IWorkloadFactory::IsLayerConfigurationSupported(const BackendId& backendId,
         }
         case LayerType::ReverseV2:
         {
-            auto cLayer = PolymorphicDowncast<const ReverseV2Layer*>(&layer);
-            const TensorInfo& input  = layer.GetInputSlot(0).GetTensorInfo();
+            const TensorInfo& input0  = layer.GetInputSlot(0).GetConnection()->GetTensorInfo();
+            const TensorInfo& input1  = layer.GetInputSlot(1).GetConnection()->GetTensorInfo();
             const TensorInfo& output = layer.GetOutputSlot(0).GetTensorInfo();
-            result = layerSupportObject.IsReverseV2Supported(OverrideDataType(input, dataType),
+            result = layerSupportObject.IsReverseV2Supported(OverrideDataType(input0, dataType),
+                                                             OverrideDataType(input1, armnn::DataType::Signed32),
                                                              OverrideDataType(output, dataType),
-                                                             cLayer->GetParameters(),
                                                              reason);
             break;
         }
diff --git a/src/backends/backendsCommon/test/IsLayerSupportedTestImpl.hpp b/src/backends/backendsCommon/test/IsLayerSupportedTestImpl.hpp
index f7a852f440..182fab97be 100644
--- a/src/backends/backendsCommon/test/IsLayerSupportedTestImpl.hpp
+++ b/src/backends/backendsCommon/test/IsLayerSupportedTestImpl.hpp
@@ -744,8 +744,6 @@ DECLARE_LAYER_POLICY_2_PARAM(Resize)
 
 DECLARE_LAYER_POLICY_2_PARAM(Reshape)
 
-DECLARE_LAYER_POLICY_2_PARAM(ReverseV2)
-
 DECLARE_LAYER_POLICY_1_PARAM(Shape)
 
 DECLARE_LAYER_POLICY_2_PARAM(Slice)
diff --git a/src/backends/backendsCommon/test/layerTests/ReverseV2TestImpl.cpp b/src/backends/backendsCommon/test/layerTests/ReverseV2TestImpl.cpp
index 586b831e45..3cfd614336 100644
--- a/src/backends/backendsCommon/test/layerTests/ReverseV2TestImpl.cpp
+++ b/src/backends/backendsCommon/test/layerTests/ReverseV2TestImpl.cpp
@@ -23,23 +23,25 @@ namespace
         armnn::IWorkloadFactory& workloadFactory,
         const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
         const armnn::ITensorHandleFactory& tensorHandleFactory,
-        armnn::ReverseV2Descriptor descriptor,
         const std::vector<T>& input,
+        const std::vector<int>& axis,
         const std::vector<T>& outputExpected,
         const armnn::TensorInfo& inputInfo,
+        const armnn::TensorInfo& axisInfo,
         const armnn::TensorInfo& outputInfo)
     {
         LayerTestResult<T, NumDims> result(outputInfo);
         std::vector<T> outputActual(outputInfo.GetNumElements());
 
         std::unique_ptr<armnn::ITensorHandle> inputHandle = tensorHandleFactory.CreateTensorHandle(inputInfo);
+        std::unique_ptr<armnn::ITensorHandle> axisHandle = tensorHandleFactory.CreateTensorHandle(axisInfo);
         std::unique_ptr<armnn::ITensorHandle> outputHandle = tensorHandleFactory.CreateTensorHandle(outputInfo);
 
         armnn::ReverseV2QueueDescriptor queueDescriptor;
-        queueDescriptor.m_Parameters = std::move(descriptor);
         armnn::WorkloadInfo workloadInfo;
 
         AddInputToWorkload(queueDescriptor, workloadInfo, inputInfo, inputHandle.get());
+        AddInputToWorkload(queueDescriptor, workloadInfo, axisInfo, axisHandle.get());
         AddOutputToWorkload(queueDescriptor, workloadInfo, outputInfo, outputHandle.get());
 
         // Don't execute if ReverseV2 is not supported, as an exception will be raised.
@@ -47,9 +49,9 @@ namespace
         std::string reasonIfUnsupported;
         armnn::LayerSupportHandle handle = armnn::GetILayerSupportByBackendId(backend);
         result.m_Supported = handle.IsReverseV2Supported(inputInfo,
-                                                           outputInfo,
-                                                           queueDescriptor.m_Parameters,
-                                                           reasonIfUnsupported);
+                                                         axisInfo,
+                                                         outputInfo,
+                                                         reasonIfUnsupported);
         if (!result.m_Supported)
         {
             return result;
@@ -58,9 +60,11 @@ namespace
         auto workload = workloadFactory.CreateWorkload(armnn::LayerType::ReverseV2, queueDescriptor, workloadInfo);
 
         inputHandle->Allocate();
+        axisHandle->Allocate();
         outputHandle->Allocate();
 
         CopyDataToITensorHandle(inputHandle.get(), input.data());
+        CopyDataToITensorHandle(axisHandle.get(), axis.data());
 
         workload->PostAllocationConfigure();
         ExecuteWorkload(*workload, memoryManager);
@@ -80,13 +84,13 @@ LayerTestResult<T, 2> ReverseV2SimpleTestEmptyAxis(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor. No axes set so output is the same as input
-    auto descriptor = armnn::ReverseV2Descriptor();
+    // Simple test with no axes set so output is the same as input
 
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2,2}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2,2}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -94,6 +98,8 @@ LayerTestResult<T, 2> ReverseV2SimpleTestEmptyAxis(
         3, 4
     }, qScale, qOffset);
 
+    std::vector<int> axis = armnnUtils::QuantizedVector<int>({}, qScale, qOffset);
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         1, 2,
         3, 4
@@ -102,10 +108,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTestEmptyAxis(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                                 memoryManager,
                                                 tensorHandleFactory,
-                                                descriptor,
                                                 input,
+                                                axis,
                                                 outputExpected,
                                                 inputInfo,
+                                                axisInfo,
                                                 outputInfo);
 }
 
@@ -115,26 +122,27 @@ LayerTestResult<T, 2> ReverseV2SimpleTestEmptyTensor(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor. Empty tensor set so output is the same as input
-    auto descriptor = armnn::ReverseV2Descriptor();
+    // Simple test with empty input tensor
 
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({0}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({0}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({0}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({}, qScale, qOffset);
-
+    std::vector<int> axis = armnnUtils::QuantizedVector<int>({}, qScale, qOffset);
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({}, qScale, qOffset);
 
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -144,12 +152,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest1Dim(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {0});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({4}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({1}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({4}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -157,6 +164,8 @@ LayerTestResult<T, 2> ReverseV2SimpleTest1Dim(
         3, 4
     }, qScale, qOffset);
 
+    std::vector<int> axis = {0};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         4, 3,
         2, 1
@@ -165,10 +174,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest1Dim(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -178,12 +188,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest2Dim1Axis(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2,2}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({1}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2,2}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -191,6 +200,8 @@ LayerTestResult<T, 2> ReverseV2SimpleTest2Dim1Axis(
         3, 4
     }, qScale, qOffset);
 
+    std::vector<int> axis = {1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         2, 1,
         4, 3
@@ -199,10 +210,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest2Dim1Axis(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -212,12 +224,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest2Dim2Axis(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {1, 0});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2,2}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({2}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2,2}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -225,6 +236,8 @@ LayerTestResult<T, 2> ReverseV2SimpleTest2Dim2Axis(
         3, 4
     }, qScale, qOffset);
 
+    std::vector<int> axis = {1,0};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         4, 3,
         2, 1
@@ -233,10 +246,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest2Dim2Axis(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -246,12 +260,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest3Dim1Axis(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2, 3, 4}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({1}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2, 3, 4}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -263,6 +276,8 @@ LayerTestResult<T, 2> ReverseV2SimpleTest3Dim1Axis(
         21, 22, 23, 24
     }, qScale, qOffset);
 
+    std::vector<int> axis = {1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         9, 10, 11, 12,
         5, 6, 7, 8,
@@ -275,10 +290,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest3Dim1Axis(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -288,12 +304,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest3Dim2Axis(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {0, 1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2, 3, 4}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({2}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2, 3, 4}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -305,6 +320,8 @@ LayerTestResult<T, 2> ReverseV2SimpleTest3Dim2Axis(
         21, 22, 23, 24
     }, qScale, qOffset);
 
+    std::vector<int> axis = {0, 1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         21, 22, 23, 24,
         17, 18, 19, 20,
@@ -317,10 +334,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest3Dim2Axis(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -330,14 +348,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest3Dim3Axis(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {1, 0, 2});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2, 3, 4}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({3}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2, 3, 4}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -349,6 +364,8 @@ LayerTestResult<T, 2> ReverseV2SimpleTest3Dim3Axis(
         21, 22, 23, 24
     }, qScale, qOffset);
 
+    std::vector<int> axis = {1, 0, 2};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         24, 23, 22, 21,
         20, 19, 18, 17,
@@ -361,10 +378,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest3Dim3Axis(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -374,14 +392,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest4Dim1Axis(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {0});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2, 2, 2, 3}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({1}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2, 2, 2, 3}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -395,6 +410,8 @@ LayerTestResult<T, 2> ReverseV2SimpleTest4Dim1Axis(
         22, 23, 24
     }, qScale, qOffset);
 
+    std::vector<int> axis = {0};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         13, 14, 15,
         16, 17, 18,
@@ -409,10 +426,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest4Dim1Axis(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -422,14 +440,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest4Dim2Axis(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {0, 1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2, 2, 2, 3}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({2}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2, 2, 2, 3}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -443,6 +458,8 @@ LayerTestResult<T, 2> ReverseV2SimpleTest4Dim2Axis(
         22, 23, 24
     }, qScale, qOffset);
 
+    std::vector<int> axis = {0, 1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         19, 20, 21,
         22, 23, 24,
@@ -457,10 +474,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest4Dim2Axis(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -470,14 +488,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest4Dim3Axis(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {0, 1, 2});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2, 2, 2, 3}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({3}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2, 2, 2, 3}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -491,6 +506,8 @@ LayerTestResult<T, 2> ReverseV2SimpleTest4Dim3Axis(
         22, 23, 24
     }, qScale, qOffset);
 
+    std::vector<int> axis = {0, 1, 2};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         22, 23, 24,
         19, 20, 21,
@@ -505,10 +522,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest4Dim3Axis(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -518,14 +536,11 @@ LayerTestResult<T, 2> ReverseV2SimpleTest4Dim4Axis(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {0, 1, 2, 3});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2, 2, 2, 3}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({4}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2, 2, 2, 3}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -539,24 +554,27 @@ LayerTestResult<T, 2> ReverseV2SimpleTest4Dim4Axis(
         22, 23, 24
     }, qScale, qOffset);
 
+    std::vector<int> axis = {0, 1, 2, 3};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
-                                                                           24, 23, 22,
-                                                                           21, 20, 19,
-                                                                           18, 17, 16,
-                                                                           15, 14, 13,
-                                                                           12, 11, 10,
-                                                                           9, 8, 7,
-                                                                           6, 5, 4,
-                                                                           3, 2, 1
-                                                                   }, qScale, qOffset);
+       24, 23, 22,
+       21, 20, 19,
+       18, 17, 16,
+       15, 14, 13,
+       12, 11, 10,
+       9, 8, 7,
+       6, 5, 4,
+       3, 2, 1
+    }, qScale, qOffset);
 
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -566,13 +584,11 @@ LayerTestResult<T, 2> ReverseV2EvenRowOddColTest2Dim(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor. No axes set so output is the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2, 3}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({1}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2, 3}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -580,6 +596,8 @@ LayerTestResult<T, 2> ReverseV2EvenRowOddColTest2Dim(
         4, 5, 6
     }, qScale, qOffset);
 
+    std::vector<int> axis = {1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         3, 2, 1,
         6, 5, 4
@@ -588,10 +606,11 @@ LayerTestResult<T, 2> ReverseV2EvenRowOddColTest2Dim(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -601,14 +620,11 @@ LayerTestResult<T, 2> ReverseV2EvenRowOddColTest3Dim(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2, 3, 1}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({1}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2, 3, 1}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -616,6 +632,8 @@ LayerTestResult<T, 2> ReverseV2EvenRowOddColTest3Dim(
         4, 5, 6
     }, qScale, qOffset);
 
+    std::vector<int> axis = {1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         3, 2, 1,
         6, 5, 4
@@ -624,10 +642,11 @@ LayerTestResult<T, 2> ReverseV2EvenRowOddColTest3Dim(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -637,14 +656,11 @@ LayerTestResult<T, 2> ReverseV2EvenRowEvenColTest2Dim(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2, 4}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({1}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2, 4}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -652,6 +668,8 @@ LayerTestResult<T, 2> ReverseV2EvenRowEvenColTest2Dim(
         5, 6, 7, 8
     }, qScale, qOffset);
 
+    std::vector<int> axis = {1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         4, 3, 2, 1,
         8, 7, 6, 5
@@ -660,10 +678,11 @@ LayerTestResult<T, 2> ReverseV2EvenRowEvenColTest2Dim(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -673,14 +692,11 @@ LayerTestResult<T, 2> ReverseV2EvenRowEvenColTest3Dim(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2, 4, 1}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({1}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2, 4, 1}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -688,6 +704,8 @@ LayerTestResult<T, 2> ReverseV2EvenRowEvenColTest3Dim(
         5, 6, 7, 8
     }, qScale, qOffset);
 
+    std::vector<int> axis = {1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         4, 3, 2, 1,
         8, 7, 6, 5
@@ -696,10 +714,11 @@ LayerTestResult<T, 2> ReverseV2EvenRowEvenColTest3Dim(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -709,14 +728,11 @@ LayerTestResult<T, 2> ReverseV2OddRowOddColTest2Dim(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({3, 3}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({1}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({3, 3}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -725,6 +741,8 @@ LayerTestResult<T, 2> ReverseV2OddRowOddColTest2Dim(
         7, 8, 9
     }, qScale, qOffset);
 
+    std::vector<int> axis = {1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         3, 2, 1,
         6, 5, 4,
@@ -734,10 +752,11 @@ LayerTestResult<T, 2> ReverseV2OddRowOddColTest2Dim(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -747,14 +766,11 @@ LayerTestResult<T, 2> ReverseV2OddRowOddColTest3Dim(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({3, 3, 1}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({1}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({3, 3, 1}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -763,6 +779,8 @@ LayerTestResult<T, 2> ReverseV2OddRowOddColTest3Dim(
         7, 8, 9
     }, qScale, qOffset);
 
+    std::vector<int> axis = {1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         3, 2, 1,
         6, 5, 4,
@@ -772,10 +790,11 @@ LayerTestResult<T, 2> ReverseV2OddRowOddColTest3Dim(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -785,14 +804,11 @@ LayerTestResult<T, 2> ReverseV2OddRowEvenColTest2Dim(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({3, 4}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({1}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({3, 4}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -801,6 +817,8 @@ LayerTestResult<T, 2> ReverseV2OddRowEvenColTest2Dim(
         9, 10, 11, 12
     }, qScale, qOffset);
 
+    std::vector<int> axis = {1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         4, 3, 2, 1,
         8, 7, 6, 5,
@@ -810,10 +828,11 @@ LayerTestResult<T, 2> ReverseV2OddRowEvenColTest2Dim(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -823,14 +842,11 @@ LayerTestResult<T, 2> ReverseV2OddRowEvenColTest3Dim(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({3, 4, 1}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({1}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({3, 4, 1}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -839,6 +855,8 @@ LayerTestResult<T, 2> ReverseV2OddRowEvenColTest3Dim(
         9, 10, 11, 12
     }, qScale, qOffset);
 
+    std::vector<int> axis = {1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         4, 3, 2, 1,
         8, 7, 6, 5,
@@ -848,10 +866,11 @@ LayerTestResult<T, 2> ReverseV2OddRowEvenColTest3Dim(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -861,14 +880,11 @@ LayerTestResult<T, 2> ReverseV2NegAxisTest2Dim1Axis(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {-1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2, 4}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({1}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2, 4}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -876,6 +892,8 @@ LayerTestResult<T, 2> ReverseV2NegAxisTest2Dim1Axis(
         5, 6, 7, 8,
     }, qScale, qOffset);
 
+    std::vector<int> axis = {-1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         4, 3, 2, 1,
         8, 7, 6, 5
@@ -884,10 +902,11 @@ LayerTestResult<T, 2> ReverseV2NegAxisTest2Dim1Axis(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -897,14 +916,11 @@ LayerTestResult<T, 2> ReverseV2NegAxisTest3Dim2Axis(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     const armnn::ITensorHandleFactory& tensorHandleFactory)
 {
-    // Simple test with default descriptor.  No axes set so output is
-    // the same as input
-    auto descriptor = armnn::ReverseV2Descriptor(std::vector<int> {1, -1});
-
     float qScale = 1.0f;
     int32_t qOffset = 0;
 
     armnn::TensorInfo inputInfo({2, 4, 1}, ArmnnType, qScale, qOffset);
+    armnn::TensorInfo axisInfo({2}, armnn::DataType::Signed32, qScale, qOffset);
     armnn::TensorInfo outputInfo({2, 4, 1}, ArmnnType, qScale, qOffset);
 
     std::vector<T> input = armnnUtils::QuantizedVector<T>({
@@ -912,6 +928,8 @@ LayerTestResult<T, 2> ReverseV2NegAxisTest3Dim2Axis(
         5, 6, 7, 8,
     }, qScale, qOffset);
 
+    std::vector<int> axis = {1, -1};
+
     std::vector<T> outputExpected = armnnUtils::QuantizedVector<T>({
         4, 3, 2, 1,
         8, 7, 6, 5
@@ -920,10 +938,11 @@ LayerTestResult<T, 2> ReverseV2NegAxisTest3Dim2Axis(
     return ReverseV2TestImpl<ArmnnType, T, 2>(workloadFactory,
                                               memoryManager,
                                               tensorHandleFactory,
-                                              descriptor,
                                               input,
+                                              axis,
                                               outputExpected,
                                               inputInfo,
+                                              axisInfo,
                                               outputInfo);
 }
 
@@ -1081,4 +1100,4 @@ template LayerTestResult<armnn::ResolveType<armnn::DataType::QSymmS16>, 2>
 ReverseV2SimpleTest2Dim2Axis<armnn::DataType::QSymmS16>(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
-    const armnn::ITensorHandleFactory& tensorHandleFactory);
\ No newline at end of file
+    const armnn::ITensorHandleFactory& tensorHandleFactory);
diff --git a/src/backends/reference/RefLayerSupport.cpp b/src/backends/reference/RefLayerSupport.cpp
index 1d5fab1adc..e94478f088 100644
--- a/src/backends/reference/RefLayerSupport.cpp
+++ b/src/backends/reference/RefLayerSupport.cpp
@@ -344,7 +344,7 @@ bool RefLayerSupport::IsLayerSupported(const LayerType& type,
         case LayerType::ReverseV2:
             return IsReverseV2Supported(infos[0],
                                         infos[1],
-                                        *(PolymorphicDowncast<const ReverseV2Descriptor*>(&descriptor)),
+                                        infos[2],
                                         reasonIfUnsupported);
         case LayerType::Reduce:
             return IsReduceSupported(infos[0],
@@ -2361,12 +2361,11 @@ bool RefLayerSupport::IsResizeSupported(const TensorInfo& input,
     return supported;
 }
 
-bool RefLayerSupport::IsReverseV2Supported(const TensorInfo& input,
+bool RefLayerSupport::IsReverseV2Supported(const TensorInfo& input0,
+                                           const TensorInfo& input1,
                                            const TensorInfo& output,
-                                           const ReverseV2Descriptor& descriptor,
                                            Optional<std::string&> reasonIfUnsupported) const
 {
-    IgnoreUnused(descriptor);
     bool supported = true;
     // ReverseV2 is data type agnostic so it can support all the types in the Reference backend
     std::array<DataType,6> supportedTypes =
@@ -2379,14 +2378,22 @@ bool RefLayerSupport::IsReverseV2Supported(const TensorInfo& input,
         DataType::QSymmS16
     };
 
-    supported &= CheckSupportRule(TypeAnyOf(input, supportedTypes), reasonIfUnsupported,
-                                  "Reference ReverseV2: input type not supported");
+    supported &= CheckSupportRule(TypeAnyOf(input0, supportedTypes), reasonIfUnsupported,
+                                  "Reference ReverseV2: input0 type not supported");
 
     supported &= CheckSupportRule(TypeAnyOf(output, supportedTypes), reasonIfUnsupported,
                                   "Reference ReverseV2: output type not supported");
 
-    supported &= CheckSupportRule(TypesAreEqual(input, output), reasonIfUnsupported,
-                                  "Reference ReverseV2: input and output types not matching");
+    supported &= CheckSupportRule(TypesAreEqual(input0, output), reasonIfUnsupported,
+                                  "Reference ReverseV2: input0 and output types not matching");
+
+    std::array<DataType,6> input2SupportedTypes =
+    {
+        DataType::Signed32
+    };
+
+    supported &= CheckSupportRule(TypeAnyOf(input1, input2SupportedTypes), reasonIfUnsupported,
+                                  "Reference ReverseV2: input1 type not supported");
 
     return supported;
 }
diff --git a/src/backends/reference/RefLayerSupport.hpp b/src/backends/reference/RefLayerSupport.hpp
index 0afb9c2c94..21d59e27fc 100644
--- a/src/backends/reference/RefLayerSupport.hpp
+++ b/src/backends/reference/RefLayerSupport.hpp
@@ -299,9 +299,9 @@ public:
                            const ResizeDescriptor& descriptor,
                            Optional<std::string&> reasonIfUnsupported = EmptyOptional()) const override;
 
-    bool IsReverseV2Supported(const TensorInfo& input,
+    bool IsReverseV2Supported(const TensorInfo& input0,
+                              const TensorInfo& input1,
                               const TensorInfo& output,
-                              const ReverseV2Descriptor& descriptor,
                               Optional<std::string&> reasonIfUnsupported = EmptyOptional()) const;
 
     bool IsShapeSupported(const TensorInfo& input,
diff --git a/src/backends/reference/workloads/RefReverseV2Workload.cpp b/src/backends/reference/workloads/RefReverseV2Workload.cpp
index cd2d9f930b..22d5449466 100644
--- a/src/backends/reference/workloads/RefReverseV2Workload.cpp
+++ b/src/backends/reference/workloads/RefReverseV2Workload.cpp
@@ -32,16 +32,21 @@ namespace armnn
         ARMNN_SCOPED_PROFILING_EVENT(Compute::CpuRef, "RefReverseV2Workload_Execute");
 
         const TensorInfo& inputInfo = GetTensorInfo(inputs[0]);
+        const TensorInfo& axisInfo = GetTensorInfo(inputs[1]);
 
         std::unique_ptr<Decoder<float>> inputDecoder = MakeDecoder<float>(GetTensorInfo(inputs[0]),
                                                                           inputs[0]->Map());
 
+        std::unique_ptr<Decoder<int>> axisDecoder = MakeDecoder<int>(GetTensorInfo(inputs[1]),
+                                                                          inputs[1]->Map());
+
         std::unique_ptr<Encoder<float>> outputEncoder = MakeEncoder<float>(GetTensorInfo(outputs[0]),
                                                                            outputs[0]->Map());
 
-        ReverseV2(m_Data.m_Parameters,
-                  inputInfo,
+        ReverseV2(inputInfo,
+                  axisInfo,
                   *inputDecoder,
+                  *axisDecoder,
                   *outputEncoder);
     }
 
diff --git a/src/backends/reference/workloads/ReverseV2Impl.cpp b/src/backends/reference/workloads/ReverseV2Impl.cpp
index f6d5fd74d1..896f9050f5 100644
--- a/src/backends/reference/workloads/ReverseV2Impl.cpp
+++ b/src/backends/reference/workloads/ReverseV2Impl.cpp
@@ -75,13 +75,16 @@ unsigned int ReverseRelocateIdx(unsigned int idx,
     return outputIdx;
 }
 
-void ReverseV2(const ReverseV2Descriptor& params,
-               const TensorInfo& inputInfo,
+void ReverseV2(const TensorInfo& inputInfo,
+               const TensorInfo& axisInfo,
                Decoder<float>& inputDecoder,
+               Decoder<int>& axisDecoder,
                Encoder<float>& outputEncoder)
 {
+    unsigned int axesRank = static_cast<unsigned int>(axisInfo.GetNumElements());
+
     // Empty axis and empty tensor case: copy input to output
-    if (params.m_Axis.empty() || inputInfo.GetNumElements() == 0)
+    if ((axesRank == 0) || inputInfo.GetNumElements() == 0)
     {
         for (unsigned idx = 0; idx < inputInfo.GetNumElements(); idx++)
         {
@@ -95,11 +98,19 @@ void ReverseV2(const ReverseV2Descriptor& params,
 
     unsigned int inputRank = static_cast<unsigned int>(inputInfo.GetNumDimensions());
 
-    std::vector<bool>axisFlag(inputRank, false);
-    std::vector<unsigned int>dimSize(inputRank, 0);
+    std::vector<bool> axisFlag(inputRank, false);
+    std::vector<unsigned int> dimSize(inputRank, 0);
+    std::vector<int32_t> axis(axesRank, 0);
+
+    // Decode the axis information
+    for (unsigned int i=0; i < axesRank; i++)
+    {
+        axis[i] = axisDecoder.Get();
+        axisDecoder += 1;
+    }
 
     // Make sure the axes are positive
-    for (int32_t axisElement: params.m_Axis)
+    for (int32_t axisElement: axis)
     {
         axisElement = axisElement < 0 ? axisElement + static_cast<int32_t>(inputRank) : axisElement;
         axisFlag[static_cast<uint32_t>(axisElement)] = true;
diff --git a/src/backends/reference/workloads/ReverseV2Impl.hpp b/src/backends/reference/workloads/ReverseV2Impl.hpp
index bc1fe1d432..59407d4a4e 100644
--- a/src/backends/reference/workloads/ReverseV2Impl.hpp
+++ b/src/backends/reference/workloads/ReverseV2Impl.hpp
@@ -13,9 +13,10 @@
 namespace armnn
 {
 
-void ReverseV2(const ReverseV2Descriptor& params,
-               const TensorInfo& inputInfo,
+void ReverseV2(const TensorInfo& inputInfo,
+               const TensorInfo& axisInfo,
                Decoder<float>& inputDecoder,
+               Decoder<int>& axisDecoder,
                Encoder<float>& outputEncoder);
 
-} // namespace armnn
\ No newline at end of file
+} // namespace armnn