IVGCVSW-6980 Delegate support for slice operator

Signed-off-by: Cathal Corbett <cathal.corbett@arm.com>
Change-Id: I90d800160b070e25d999b5102a7ce6d3e0ed6a81

9 files changed, 729 insertions, 353 deletions

diff --git a/delegate/CMakeLists.txt b/delegate/CMakeLists.txt
index 847a8a0be5..fe5c962321 100644
--- a/delegate/CMakeLists.txt
+++ b/delegate/CMakeLists.txt
@@ -202,6 +202,8 @@ if(BUILD_UNIT_TESTS)
         src/test/ShapeTestHelper.hpp
         src/test/SliceTest.cpp
         src/test/SliceTestHelper.hpp
+        src/test/StridedSliceTest.cpp
+        src/test/StridedSliceTestHelper.hpp
         src/test/SplitTest.cpp
         src/test/SplitTestHelper.hpp
         src/test/TestUtils.hpp
diff --git a/delegate/src/Slice.hpp b/delegate/src/Slice.hpp
index 6e355ae741..cbcb45ec65 100644
--- a/delegate/src/Slice.hpp
+++ b/delegate/src/Slice.hpp
@@ -1,5 +1,5 @@
 //
-// Copyright © 2020 Arm Ltd and Contributors. All rights reserved.
+// Copyright © 2022 Arm Ltd and Contributors. All rights reserved.
 // SPDX-License-Identifier: MIT
 //
 
@@ -21,10 +21,10 @@ TfLiteStatus VisitSliceOperator(DelegateData& delegateData,
                                 int nodeIndex,
                                 int32_t sliceOperatorCode)
 {
-    TF_LITE_ENSURE_STATUS(ValidateNumInputs(tfLiteContext, tfLiteNode, 4, nodeIndex));
+    TF_LITE_ENSURE_STATUS(ValidateNumInputs(tfLiteContext, tfLiteNode, 3, nodeIndex));
     TF_LITE_ENSURE_STATUS(ValidateNumOutputs(tfLiteContext, tfLiteNode, 1, nodeIndex));
 
-    // Read inputs [input, begin, end, strides]
+    // Read inputs [input, begin, size]
     int numInputs = tfLiteNode->inputs->size;
     std::vector<const TfLiteTensor*> tfLiteInputs;
     tfLiteInputs.reserve(numInputs);
@@ -39,15 +39,15 @@ TfLiteStatus VisitSliceOperator(DelegateData& delegateData,
         }
     }
 
-    // We save the begin, end and strides tensors in our descriptor. Therefore we have to read those values from inputs
+    // We save the begin and size tensors in our descriptor. Therefore we have to read those values from inputs
     int inputRank = tfLiteInputs[0]->dims->size;
-    auto ReadInt32Input = [&](int inputIndex, std::vector<int32_t>& outputData) ->  TfLiteStatus
+    auto ReadInt32Input = [&](int inputIndex, std::vector<uint32_t>& outputData) ->  TfLiteStatus
     {
         if (tfLiteInputs[inputIndex]->type != kTfLiteInt32)
         {
             TF_LITE_MAYBE_KERNEL_LOG(
                     tfLiteContext,
-                    "TfLiteArmnnDelegate: The Begin-, End- and Stride-Tensors of the StridedSlice operation need to "
+                    "TfLiteArmnnDelegate: The Begin- and Size-Tensors of the Slice operation need to "
                     "be of type int32. Operator: #%d node #%d: ",
                     sliceOperatorCode, nodeIndex);
             return kTfLiteError;
@@ -57,7 +57,7 @@ TfLiteStatus VisitSliceOperator(DelegateData& delegateData,
         {
             TF_LITE_MAYBE_KERNEL_LOG(
                     tfLiteContext,
-                    "TfLiteArmnnDelegate: The Begin-, End- and Stride-Tensors of the StridedSlice operation need to "
+                    "TfLiteArmnnDelegate: The Begin- and Size-Tensors of the Slice operation need to "
                     "be a 1D-Tensor. Operator: #%d node #%d: ",
                     sliceOperatorCode, nodeIndex);
             return kTfLiteError;
@@ -67,41 +67,26 @@ TfLiteStatus VisitSliceOperator(DelegateData& delegateData,
         {
             TF_LITE_MAYBE_KERNEL_LOG(
                     tfLiteContext,
-                    "TfLiteArmnnDelegate: The number of values in the Begin-, End- and Stride-Tensors of the "
-                    "StridedSlice operation need to be equal to the rank of the Input-Tensor. Operator: #%d node #%d: ",
+                    "TfLiteArmnnDelegate: The number of values in the Begin- and Size-Tensors of the "
+                    "Slice operation need to be equal to the rank of the Input-Tensor. Operator: #%d node #%d: ",
                     sliceOperatorCode, nodeIndex);
             return kTfLiteError;
         }
         // return tensor data
-        auto* tensorDataPtr = tflite::GetTensorData<int32_t>(tfLiteInputs[inputIndex]);
+        auto* tensorDataPtr = tflite::GetTensorData<uint32_t>(tfLiteInputs[inputIndex]);
         outputData.assign(tensorDataPtr, tensorDataPtr+numValues);
         return kTfLiteOk;
     };
 
-    std::vector<int32_t> beginData;
-    if (ReadInt32Input(1, beginData) != kTfLiteOk)
+    std::vector<uint32_t> begin;
+    if (ReadInt32Input(1, begin) != kTfLiteOk)
         return kTfLiteError;
-    std::vector<int32_t> endData;
-    if (ReadInt32Input(2, endData) != kTfLiteOk)
+    std::vector<uint32_t> size;
+    if (ReadInt32Input(2, size) != kTfLiteOk)
         return kTfLiteError;
-    std::vector<int32_t> strideData;
-    if (ReadInt32Input(3, strideData) != kTfLiteOk)
-        return kTfLiteError;
-
-    // parse built in options
-    auto* stridedSliceParams = reinterpret_cast<TfLiteStridedSliceParams*>(tfLiteNode->builtin_data);
 
     // Write all data to the descriptor
-    armnn::StridedSliceDescriptor descriptor;
-    descriptor.m_Begin          = std::move(beginData);
-    descriptor.m_End            = std::move(endData);
-    descriptor.m_Stride         = std::move(strideData);
-    descriptor.m_BeginMask      = stridedSliceParams->begin_mask;
-    descriptor.m_EllipsisMask   = stridedSliceParams->ellipsis_mask;
-    descriptor.m_EndMask        = stridedSliceParams->end_mask;
-    descriptor.m_NewAxisMask    = stridedSliceParams->new_axis_mask;
-    descriptor.m_ShrinkAxisMask = stridedSliceParams->shrink_axis_mask;
-    descriptor.m_DataLayout     = armnn::DataLayout::NHWC;
+    armnn::SliceDescriptor descriptor(begin, size);
 
     // Validate output
     const TfLiteTensor& tfLiteOutputTensor = tfLiteTensors[tfLiteNode->outputs->data[0]];
@@ -118,7 +103,7 @@ TfLiteStatus VisitSliceOperator(DelegateData& delegateData,
     {
         FORWARD_LAYER_SUPPORT_FUNC("SLICE",
                                    tfLiteContext,
-                                   IsStridedSliceSupported,
+                                   IsSliceSupported,
                                    delegateData.m_Backends,
                                    isSupported,
                                    inputTensorInfo,
@@ -133,7 +118,7 @@ TfLiteStatus VisitSliceOperator(DelegateData& delegateData,
     }
 
     // Add a StridedSlice layer
-    armnn::IConnectableLayer* layer = delegateData.m_Network->AddStridedSliceLayer(descriptor);
+    armnn::IConnectableLayer* layer = delegateData.m_Network->AddSliceLayer(descriptor);
     ARMNN_ASSERT(layer != nullptr);
 
     armnn::IOutputSlot& outputSlot = layer->GetOutputSlot(0);
diff --git a/delegate/src/StridedSlice.hpp b/delegate/src/StridedSlice.hpp
new file mode 100644
index 0000000000..515c819ffe
--- /dev/null
+++ b/delegate/src/StridedSlice.hpp
@@ -0,0 +1,146 @@
+//
+// Copyright © 2022 Arm Ltd and Contributors. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include <armnn/utility/IgnoreUnused.hpp>
+
+#include <tensorflow/lite/builtin_ops.h>
+#include <tensorflow/lite/c/builtin_op_data.h>
+#include <tensorflow/lite/c/common.h>
+#include <tensorflow/lite/minimal_logging.h>
+
+namespace armnnDelegate
+{
+
+TfLiteStatus VisitStridedSliceOperator(DelegateData& delegateData,
+                                       TfLiteContext* tfLiteContext,
+                                       TfLiteNode* tfLiteNode,
+                                       int nodeIndex,
+                                       int32_t sliceOperatorCode)
+{
+    TF_LITE_ENSURE_STATUS(ValidateNumInputs(tfLiteContext, tfLiteNode, 4, nodeIndex));
+    TF_LITE_ENSURE_STATUS(ValidateNumOutputs(tfLiteContext, tfLiteNode, 1, nodeIndex));
+
+    // Read inputs [input, begin, end, strides]
+    int numInputs = tfLiteNode->inputs->size;
+    std::vector<const TfLiteTensor*> tfLiteInputs;
+    tfLiteInputs.reserve(numInputs);
+    const TfLiteTensor* tfLiteTensors = tfLiteContext->tensors;
+    for (int i = 0; i < numInputs; i++)
+    {
+        const TfLiteTensor* inputTensor = &tfLiteTensors[tfLiteNode->inputs->data[i]];
+        tfLiteInputs.push_back(inputTensor);
+        if (!IsValid(tfLiteContext, *inputTensor, sliceOperatorCode, nodeIndex))
+        {
+            return kTfLiteError;
+        }
+    }
+
+    // We save the begin, end and strides tensors in our descriptor. Therefore we have to read those values from inputs
+    int inputRank = tfLiteInputs[0]->dims->size;
+    auto ReadInt32Input = [&](int inputIndex, std::vector<int32_t>& outputData) ->  TfLiteStatus
+    {
+        if (tfLiteInputs[inputIndex]->type != kTfLiteInt32)
+        {
+            TF_LITE_MAYBE_KERNEL_LOG(
+                tfLiteContext,
+                "TfLiteArmnnDelegate: The Begin-, End- and Stride-Tensors of the StridedSlice operation need to "
+                "be of type int32. Operator: #%d node #%d: ",
+                sliceOperatorCode, nodeIndex);
+            return kTfLiteError;
+        }
+        int rank = tfLiteInputs[inputIndex]->dims->size;
+        if (rank != 1)
+        {
+            TF_LITE_MAYBE_KERNEL_LOG(
+                tfLiteContext,
+                "TfLiteArmnnDelegate: The Begin-, End- and Stride-Tensors of the StridedSlice operation need to "
+                "be a 1D-Tensor. Operator: #%d node #%d: ",
+                sliceOperatorCode, nodeIndex);
+            return kTfLiteError;
+        }
+        int numValues = tfLiteInputs[inputIndex]->dims->data[0];
+        if (numValues != inputRank)
+        {
+            TF_LITE_MAYBE_KERNEL_LOG(
+                tfLiteContext,
+                "TfLiteArmnnDelegate: The number of values in the Begin-, End- and Stride-Tensors of the "
+                "StridedSlice operation need to be equal to the rank of the Input-Tensor. Operator: #%d node #%d: ",
+                sliceOperatorCode, nodeIndex);
+            return kTfLiteError;
+        }
+        // return tensor data
+        auto* tensorDataPtr = tflite::GetTensorData<int32_t>(tfLiteInputs[inputIndex]);
+        outputData.assign(tensorDataPtr, tensorDataPtr+numValues);
+        return kTfLiteOk;
+    };
+
+    std::vector<int32_t> beginData;
+    if (ReadInt32Input(1, beginData) != kTfLiteOk)
+        return kTfLiteError;
+    std::vector<int32_t> endData;
+    if (ReadInt32Input(2, endData) != kTfLiteOk)
+        return kTfLiteError;
+    std::vector<int32_t> strideData;
+    if (ReadInt32Input(3, strideData) != kTfLiteOk)
+        return kTfLiteError;
+
+    // parse built in options
+    auto* stridedSliceParams = reinterpret_cast<TfLiteStridedSliceParams*>(tfLiteNode->builtin_data);
+
+    // Write all data to the descriptor
+    armnn::StridedSliceDescriptor descriptor;
+    descriptor.m_Begin          = std::move(beginData);
+    descriptor.m_End            = std::move(endData);
+    descriptor.m_Stride         = std::move(strideData);
+    descriptor.m_BeginMask      = stridedSliceParams->begin_mask;
+    descriptor.m_EllipsisMask   = stridedSliceParams->ellipsis_mask;
+    descriptor.m_EndMask        = stridedSliceParams->end_mask;
+    descriptor.m_NewAxisMask    = stridedSliceParams->new_axis_mask;
+    descriptor.m_ShrinkAxisMask = stridedSliceParams->shrink_axis_mask;
+    descriptor.m_DataLayout     = armnn::DataLayout::NHWC;
+
+    // Validate output
+    const TfLiteTensor& tfLiteOutputTensor = tfLiteTensors[tfLiteNode->outputs->data[0]];
+    if (!IsValid(tfLiteContext, tfLiteOutputTensor, sliceOperatorCode, nodeIndex))
+    {
+        return kTfLiteError;
+    }
+
+    const armnn::TensorInfo& inputTensorInfo  = GetTensorInfoForTfLiteTensor(*tfLiteInputs[0]);
+    const armnn::TensorInfo& outputTensorInfo = GetTensorInfoForTfLiteTensor(tfLiteOutputTensor);
+
+    bool isSupported = false;
+    auto validateFunc = [&](const armnn::TensorInfo& outInfo, bool& isSupported)
+    {
+        FORWARD_LAYER_SUPPORT_FUNC("STRIDED_SLICE",
+                                   tfLiteContext,
+                                   IsStridedSliceSupported,
+                                   delegateData.m_Backends,
+                                   isSupported,
+                                   inputTensorInfo,
+                                   outInfo,
+                                   descriptor);
+    };
+
+    if (!delegateData.m_Network)
+    {
+        validateFunc(outputTensorInfo, isSupported);
+        return isSupported ? kTfLiteOk : kTfLiteError;
+    }
+
+    // Add a StridedSlice layer
+    armnn::IConnectableLayer* layer = delegateData.m_Network->AddStridedSliceLayer(descriptor);
+    ARMNN_ASSERT(layer != nullptr);
+
+    armnn::IOutputSlot& outputSlot = layer->GetOutputSlot(0);
+    outputSlot.SetTensorInfo(outputTensorInfo);
+
+    // Connect
+    return Connect(layer, tfLiteNode, delegateData);
+}
+
+} // namespace armnnDelegate
diff --git a/delegate/src/armnn_delegate.cpp b/delegate/src/armnn_delegate.cpp
index b2ad16051f..4d95522dbd 100644
--- a/delegate/src/armnn_delegate.cpp
+++ b/delegate/src/armnn_delegate.cpp
@@ -34,6 +34,7 @@
 #include "Round.hpp"
 #include "Shape.hpp"
 #include "Slice.hpp"
+#include "StridedSlice.hpp"
 #include "Softmax.hpp"
 #include "SpaceDepth.hpp"
 #include "Split.hpp"
@@ -963,12 +964,18 @@ TfLiteStatus ArmnnSubgraph::VisitNode(DelegateData& delegateData,
                                         tfLiteNode,
                                         nodeIndex,
                                         kTfLiteBuiltinSqueeze);
-        case kTfLiteBuiltinStridedSlice:
+        case kTfLiteBuiltinSlice:
             return VisitSliceOperator(delegateData,
                                       tfLiteContext,
                                       tfLiteNode,
                                       nodeIndex,
-                                      kTfLiteBuiltinStridedSlice);
+                                      kTfLiteBuiltinSlice);
+        case kTfLiteBuiltinStridedSlice:
+            return VisitStridedSliceOperator(delegateData,
+                                             tfLiteContext,
+                                             tfLiteNode,
+                                             nodeIndex,
+                                             kTfLiteBuiltinStridedSlice);
         case kTfLiteBuiltinSum:
             return VisitReduceOperator(delegateData,
                                        tfLiteContext,
diff --git a/delegate/src/test/SliceTest.cpp b/delegate/src/test/SliceTest.cpp
index bd0584936e..1d7133f1fd 100644
--- a/delegate/src/test/SliceTest.cpp
+++ b/delegate/src/test/SliceTest.cpp
@@ -1,5 +1,5 @@
 //
-// Copyright © 2021 Arm Ltd and Contributors. All rights reserved.
+// Copyright © 2022 Arm Ltd and Contributors. All rights reserved.
 // SPDX-License-Identifier: MIT
 //
 
@@ -8,236 +8,74 @@
 #include <armnn_delegate.hpp>
 
 #include <flatbuffers/flatbuffers.h>
-#include <tensorflow/lite/schema/schema_generated.h>
 
 #include <doctest/doctest.h>
 
 namespace armnnDelegate
 {
 
-void StridedSlice4DTest(std::vector<armnn::BackendId>& backends)
+void SliceFixtureSimpleTest(std::vector<armnn::BackendId>& backends)
 {
-    std::vector<int32_t> inputShape  { 3, 2, 3, 1 };
-    std::vector<int32_t> outputShape { 1, 2, 3, 1 };
-    std::vector<int32_t> beginShape  { 4 };
-    std::vector<int32_t> endShape    { 4 };
-    std::vector<int32_t> strideShape { 4 };
-
-    std::vector<int32_t> beginData  { 1, 0, 0, 0 };
-    std::vector<int32_t> endData    { 2, 2, 3, 1 };
-    std::vector<int32_t> strideData { 1, 1, 1, 1 };
+    std::vector<int32_t> inputShape  { 3, 2, 3 };
+    std::vector<int32_t> outputShape { 2, 1, 3 };
+    std::vector<int32_t> beginShape  { 3 };
+    std::vector<int32_t> sizeShape   { 3 };
+
+    std::vector<int32_t> beginData { 1, 0, 0 };
+    std::vector<int32_t> sizeData  { 2, 1, 3 };
     std::vector<float> inputData  { 1.0f, 1.0f, 1.0f, 2.0f, 2.0f, 2.0f,
                                     3.0f, 3.0f, 3.0f, 4.0f, 4.0f, 4.0f,
                                     5.0f, 5.0f, 5.0f, 6.0f, 6.0f, 6.0f };
-    std::vector<float> outputData { 3.0f, 3.0f, 3.0f, 4.0f, 4.0f, 4.0f };
-
-    StridedSliceTestImpl<float>(
-            backends,
-            inputData,
-            outputData,
-            beginData,
-            endData,
-            strideData,
-            inputShape,
-            beginShape,
-            endShape,
-            strideShape,
-            outputShape
-            );
-}
-
-void StridedSlice4DReverseTest(std::vector<armnn::BackendId>& backends)
-{
-    std::vector<int32_t> inputShape  { 3, 2, 3, 1 };
-    std::vector<int32_t> outputShape { 1, 2, 3, 1 };
-    std::vector<int32_t> beginShape  { 4 };
-    std::vector<int32_t> endShape    { 4 };
-    std::vector<int32_t> strideShape { 4 };
-
-    std::vector<int32_t> beginData  { 1, -1, 0, 0 };
-    std::vector<int32_t> endData    { 2, -3, 3, 1 };
-    std::vector<int32_t> strideData { 1, -1, 1, 1 };
-    std::vector<float>   inputData  { 1.0f, 1.0f, 1.0f, 2.0f, 2.0f, 2.0f,
-                                      3.0f, 3.0f, 3.0f, 4.0f, 4.0f, 4.0f,
-                                      5.0f, 5.0f, 5.0f, 6.0f, 6.0f, 6.0f };
-    std::vector<float>   outputData { 4.0f, 4.0f, 4.0f, 3.0f, 3.0f, 3.0f };
-
-    StridedSliceTestImpl<float>(
-            backends,
-            inputData,
-            outputData,
-            beginData,
-            endData,
-            strideData,
-            inputShape,
-            beginShape,
-            endShape,
-            strideShape,
-            outputShape
-    );
-}
+    std::vector<float> outputData { 3.0f, 3.0f, 3.0f,
+                                    5.0f, 5.0f, 5.0f };
 
-void StridedSliceSimpleStrideTest(std::vector<armnn::BackendId>& backends)
-{
-    std::vector<int32_t> inputShape  { 3, 2, 3, 1 };
-    std::vector<int32_t> outputShape { 2, 1, 2, 1 };
-    std::vector<int32_t> beginShape  { 4 };
-    std::vector<int32_t> endShape    { 4 };
-    std::vector<int32_t> strideShape { 4 };
-
-    std::vector<int32_t> beginData  { 0, 0, 0, 0 };
-    std::vector<int32_t> endData    { 3, 2, 3, 1 };
-    std::vector<int32_t> strideData { 2, 2, 2, 1 };
-    std::vector<float>   inputData  { 1.0f, 1.0f, 1.0f, 2.0f, 2.0f, 2.0f,
-                                      3.0f, 3.0f, 3.0f, 4.0f, 4.0f, 4.0f,
-                                      5.0f, 5.0f, 5.0f, 6.0f, 6.0f, 6.0f };
-    std::vector<float>   outputData { 1.0f, 1.0f,
-                                      5.0f, 5.0f };
-
-    StridedSliceTestImpl<float>(
-            backends,
-            inputData,
-            outputData,
-            beginData,
-            endData,
-            strideData,
-            inputShape,
-            beginShape,
-            endShape,
-            strideShape,
-            outputShape
-    );
+    SliceTestImpl<float>(
+        backends,
+        inputData,
+        outputData,
+        beginData,
+        sizeData,
+        inputShape,
+        beginShape,
+        sizeShape,
+        outputShape);
 }
 
-void StridedSliceSimpleRangeMaskTest(std::vector<armnn::BackendId>& backends)
+TEST_SUITE("Slice_CpuRefTests")
 {
-    std::vector<int32_t> inputShape  { 3, 2, 3, 1 };
-    std::vector<int32_t> outputShape { 3, 2, 3, 1 };
-    std::vector<int32_t> beginShape  { 4 };
-    std::vector<int32_t> endShape    { 4 };
-    std::vector<int32_t> strideShape { 4 };
-
-    std::vector<int32_t> beginData  { 1, 1, 1, 1 };
-    std::vector<int32_t> endData    { 1, 1, 1, 1 };
-    std::vector<int32_t> strideData { 1, 1, 1, 1 };
-
-    int beginMask = -1;
-    int endMask   = -1;
-
-    std::vector<float>   inputData  { 1.0f, 1.0f, 1.0f, 2.0f, 2.0f, 2.0f,
-                                      3.0f, 3.0f, 3.0f, 4.0f, 4.0f, 4.0f,
-                                      5.0f, 5.0f, 5.0f, 6.0f, 6.0f, 6.0f };
-    std::vector<float>   outputData { 1.0f, 1.0f, 1.0f, 2.0f, 2.0f, 2.0f,
-                                      3.0f, 3.0f, 3.0f, 4.0f, 4.0f, 4.0f,
-                                      5.0f, 5.0f, 5.0f, 6.0f, 6.0f, 6.0f };
-
-    StridedSliceTestImpl<float>(
-            backends,
-            inputData,
-            outputData,
-            beginData,
-            endData,
-            strideData,
-            inputShape,
-            beginShape,
-            endShape,
-            strideShape,
-            outputShape,
-            beginMask,
-            endMask
-    );
-}
-
-
-TEST_SUITE("StridedSlice_CpuRefTests")
-{
-
-TEST_CASE ("StridedSlice_4D_CpuRef_Test")
-{
-    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuRef};
-    StridedSlice4DTest(backends);
-}
-
-TEST_CASE ("StridedSlice_4D_Reverse_CpuRef_Test")
-{
-    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuRef};
-    StridedSlice4DReverseTest(backends);
-}
-
-TEST_CASE ("StridedSlice_SimpleStride_CpuRef_Test")
-{
-    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuRef};
-    StridedSliceSimpleStrideTest(backends);
-}
 
-TEST_CASE ("StridedSlice_SimpleRange_CpuRef_Test")
+TEST_CASE ("Slice_Simple_CpuRef_Test")
 {
     std::vector<armnn::BackendId> backends = {armnn::Compute::CpuRef};
-    StridedSliceSimpleRangeMaskTest(backends);
+    SliceFixtureSimpleTest(backends);
 }
 
-} // StridedSlice_CpuRefTests TestSuite
-
+} // Slice_CpuRefTests TestSuite
 
 
-TEST_SUITE("StridedSlice_CpuAccTests")
-{
 
-TEST_CASE ("StridedSlice_4D_CpuAcc_Test")
+TEST_SUITE("Slice_CpuAccTests")
 {
-    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuAcc};
-    StridedSlice4DTest(backends);
-}
 
-TEST_CASE ("StridedSlice_4D_Reverse_CpuAcc_Test")
+TEST_CASE ("Slice_Simple_CpuAcc_Test")
 {
-    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuAcc};
-    StridedSlice4DReverseTest(backends);
-}
-
-TEST_CASE ("StridedSlice_SimpleStride_CpuAcc_Test")
-{
-    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuAcc};
-    StridedSliceSimpleStrideTest(backends);
-}
-
-TEST_CASE ("StridedSlice_SimpleRange_CpuAcc_Test")
-{
-    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuAcc};
-    StridedSliceSimpleRangeMaskTest(backends);
+    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuRef};
+    SliceFixtureSimpleTest(backends);
 }
 
-} // StridedSlice_CpuAccTests TestSuite
+} // Slice_CpuAccTests TestSuite
 
 
 
 TEST_SUITE("StridedSlice_GpuAccTests")
 {
 
-TEST_CASE ("StridedSlice_4D_GpuAcc_Test")
+TEST_CASE ("Slice_Simple_GpuAcc_Test")
 {
-    std::vector<armnn::BackendId> backends = {armnn::Compute::GpuAcc};
-    StridedSlice4DTest(backends);
-}
-
-TEST_CASE ("StridedSlice_4D_Reverse_GpuAcc_Test")
-{
-    std::vector<armnn::BackendId> backends = {armnn::Compute::GpuAcc};
-    StridedSlice4DReverseTest(backends);
-}
-
-TEST_CASE ("StridedSlice_SimpleStride_GpuAcc_Test")
-{
-    std::vector<armnn::BackendId> backends = {armnn::Compute::GpuAcc};
-    StridedSliceSimpleStrideTest(backends);
-}
-
-TEST_CASE ("StridedSlice_SimpleRange_GpuAcc_Test")
-{
-    std::vector<armnn::BackendId> backends = {armnn::Compute::GpuAcc};
-    StridedSliceSimpleRangeMaskTest(backends);
+    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuRef};
+    SliceFixtureSimpleTest(backends);
 }
 
-} // StridedSlice_GpuAccTests TestSuite
+} // Slice_GpuAccTests TestSuite
 
 } // namespace armnnDelegate
\ No newline at end of file
diff --git a/delegate/src/test/SliceTestHelper.hpp b/delegate/src/test/SliceTestHelper.hpp
index abaa807aed..4a2537feec 100644
--- a/delegate/src/test/SliceTestHelper.hpp
+++ b/delegate/src/test/SliceTestHelper.hpp
@@ -1,5 +1,5 @@
 //
-// Copyright © 2021 Arm Ltd and Contributors. All rights reserved.
+// Copyright © 2022 Arm Ltd and Contributors. All rights reserved.
 // SPDX-License-Identifier: MIT
 //
 
@@ -24,61 +24,27 @@
 namespace
 {
 
-struct StridedSliceParams
-{
-    StridedSliceParams(std::vector<int32_t>& inputTensorShape,
-                       std::vector<int32_t>& beginTensorData,
-                       std::vector<int32_t>& endTensorData,
-                       std::vector<int32_t>& strideTensorData,
-                       std::vector<int32_t>& outputTensorShape,
-                       armnn::StridedSliceDescriptor& descriptor)
-        : m_InputTensorShape(inputTensorShape),
-          m_BeginTensorData(beginTensorData),
-          m_EndTensorData(endTensorData),
-          m_StrideTensorData(strideTensorData),
-          m_OutputTensorShape(outputTensorShape),
-          m_Descriptor (descriptor) {}
-
-    std::vector<int32_t> m_InputTensorShape;
-    std::vector<int32_t> m_BeginTensorData;
-    std::vector<int32_t> m_EndTensorData;
-    std::vector<int32_t> m_StrideTensorData;
-    std::vector<int32_t> m_OutputTensorShape;
-    armnn::StridedSliceDescriptor m_Descriptor;
-};
-
 std::vector<char> CreateSliceTfLiteModel(tflite::TensorType tensorType,
                                          const std::vector<int32_t>& inputTensorShape,
                                          const std::vector<int32_t>& beginTensorData,
-                                         const std::vector<int32_t>& endTensorData,
-                                         const std::vector<int32_t>& strideTensorData,
+                                         const std::vector<int32_t>& sizeTensorData,
                                          const std::vector<int32_t>& beginTensorShape,
-                                         const std::vector<int32_t>& endTensorShape,
-                                         const std::vector<int32_t>& strideTensorShape,
-                                         const std::vector<int32_t>& outputTensorShape,
-                                         const int32_t beginMask,
-                                         const int32_t endMask,
-                                         const int32_t ellipsisMask,
-                                         const int32_t newAxisMask,
-                                         const int32_t ShrinkAxisMask,
-                                         const armnn::DataLayout& dataLayout)
+                                         const std::vector<int32_t>& sizeTensorShape,
+                                         const std::vector<int32_t>& outputTensorShape)
 {
     using namespace tflite;
     flatbuffers::FlatBufferBuilder flatBufferBuilder;
 
-    std::array<flatbuffers::Offset<tflite::Buffer>, 4> buffers;
+    std::array<flatbuffers::Offset<tflite::Buffer>, 3> buffers;
     buffers[0] = CreateBuffer(flatBufferBuilder, flatBufferBuilder.CreateVector({}));
     buffers[1] = CreateBuffer(flatBufferBuilder,
                               flatBufferBuilder.CreateVector(reinterpret_cast<const uint8_t*>(beginTensorData.data()),
                                                              sizeof(int32_t) * beginTensorData.size()));
     buffers[2] = CreateBuffer(flatBufferBuilder,
-                              flatBufferBuilder.CreateVector(reinterpret_cast<const uint8_t*>(endTensorData.data()),
-                                                             sizeof(int32_t) * endTensorData.size()));
-    buffers[3] = CreateBuffer(flatBufferBuilder,
-                              flatBufferBuilder.CreateVector(reinterpret_cast<const uint8_t*>(strideTensorData.data()),
-                                                             sizeof(int32_t) * strideTensorData.size()));
+                              flatBufferBuilder.CreateVector(reinterpret_cast<const uint8_t*>(sizeTensorData.data()),
+                                                             sizeof(int32_t) * sizeTensorData.size()));
 
-    std::array<flatbuffers::Offset<Tensor>, 5> tensors;
+    std::array<flatbuffers::Offset<Tensor>, 4> tensors;
     tensors[0] = CreateTensor(flatBufferBuilder,
                               flatBufferBuilder.CreateVector<int32_t>(inputTensorShape.data(),
                                                                       inputTensorShape.size()),
@@ -92,18 +58,12 @@ std::vector<char> CreateSliceTfLiteModel(tflite::TensorType tensorType,
                               1,
                               flatBufferBuilder.CreateString("begin_tensor"));
     tensors[2] = CreateTensor(flatBufferBuilder,
-                              flatBufferBuilder.CreateVector<int32_t>(endTensorShape.data(),
-                                                                      endTensorShape.size()),
+                              flatBufferBuilder.CreateVector<int32_t>(sizeTensorShape.data(),
+                                                                      sizeTensorShape.size()),
                               ::tflite::TensorType_INT32,
                               2,
-                              flatBufferBuilder.CreateString("end_tensor"));
+                              flatBufferBuilder.CreateString("size_tensor"));
     tensors[3] = CreateTensor(flatBufferBuilder,
-                              flatBufferBuilder.CreateVector<int32_t>(strideTensorShape.data(),
-                                                                      strideTensorShape.size()),
-                              ::tflite::TensorType_INT32,
-                              3,
-                              flatBufferBuilder.CreateString("stride_tensor"));
-    tensors[4] = CreateTensor(flatBufferBuilder,
                               flatBufferBuilder.CreateVector<int32_t>(outputTensorShape.data(),
                                                                       outputTensorShape.size()),
                               tensorType,
@@ -112,45 +72,40 @@ std::vector<char> CreateSliceTfLiteModel(tflite::TensorType tensorType,
 
 
     // create operator
-    tflite::BuiltinOptions operatorBuiltinOptionsType = tflite::BuiltinOptions_StridedSliceOptions;
-    flatbuffers::Offset<void> operatorBuiltinOptions = CreateStridedSliceOptions(flatBufferBuilder,
-                                                                                 beginMask,
-                                                                                 endMask,
-                                                                                 ellipsisMask,
-                                                                                 newAxisMask,
-                                                                                 ShrinkAxisMask).Union();
-
-    const std::vector<int> operatorInputs{ 0, 1, 2, 3 };
-    const std::vector<int> operatorOutputs{ 4 };
+    tflite::BuiltinOptions operatorBuiltinOptionsType = tflite::BuiltinOptions_SliceOptions;
+    flatbuffers::Offset<void> operatorBuiltinOptions = CreateSliceOptions(flatBufferBuilder).Union();
+
+    const std::vector<int> operatorInputs{ 0, 1, 2 };
+    const std::vector<int> operatorOutputs{ 3 };
     flatbuffers::Offset <Operator> sliceOperator =
-            CreateOperator(flatBufferBuilder,
-                           0,
-                           flatBufferBuilder.CreateVector<int32_t>(operatorInputs.data(), operatorInputs.size()),
-                           flatBufferBuilder.CreateVector<int32_t>(operatorOutputs.data(), operatorOutputs.size()),
-                           operatorBuiltinOptionsType,
-                           operatorBuiltinOptions);
-
-    const std::vector<int> subgraphInputs{ 0, 1, 2, 3 };
-    const std::vector<int> subgraphOutputs{ 4 };
+        CreateOperator(flatBufferBuilder,
+                       0,
+                       flatBufferBuilder.CreateVector<int32_t>(operatorInputs.data(), operatorInputs.size()),
+                       flatBufferBuilder.CreateVector<int32_t>(operatorOutputs.data(), operatorOutputs.size()),
+                       operatorBuiltinOptionsType,
+                       operatorBuiltinOptions);
+
+    const std::vector<int> subgraphInputs{ 0, 1, 2 };
+    const std::vector<int> subgraphOutputs{ 3 };
     flatbuffers::Offset <SubGraph> subgraph =
-            CreateSubGraph(flatBufferBuilder,
-                           flatBufferBuilder.CreateVector(tensors.data(), tensors.size()),
-                           flatBufferBuilder.CreateVector<int32_t>(subgraphInputs.data(), subgraphInputs.size()),
-                           flatBufferBuilder.CreateVector<int32_t>(subgraphOutputs.data(), subgraphOutputs.size()),
-                           flatBufferBuilder.CreateVector(&sliceOperator, 1));
+        CreateSubGraph(flatBufferBuilder,
+                       flatBufferBuilder.CreateVector(tensors.data(), tensors.size()),
+                       flatBufferBuilder.CreateVector<int32_t>(subgraphInputs.data(), subgraphInputs.size()),
+                       flatBufferBuilder.CreateVector<int32_t>(subgraphOutputs.data(), subgraphOutputs.size()),
+                       flatBufferBuilder.CreateVector(&sliceOperator, 1));
 
     flatbuffers::Offset <flatbuffers::String> modelDescription =
-            flatBufferBuilder.CreateString("ArmnnDelegate: StridedSlice Operator Model");
+        flatBufferBuilder.CreateString("ArmnnDelegate: Slice Operator Model");
     flatbuffers::Offset <OperatorCode> operatorCode = CreateOperatorCode(flatBufferBuilder,
-                                                                         BuiltinOperator_STRIDED_SLICE);
+                                                                         BuiltinOperator_SLICE);
 
     flatbuffers::Offset <Model> flatbufferModel =
-            CreateModel(flatBufferBuilder,
-                        TFLITE_SCHEMA_VERSION,
-                        flatBufferBuilder.CreateVector(&operatorCode, 1),
-                        flatBufferBuilder.CreateVector(&subgraph, 1),
-                        modelDescription,
-                        flatBufferBuilder.CreateVector(buffers.data(), buffers.size()));
+        CreateModel(flatBufferBuilder,
+                    TFLITE_SCHEMA_VERSION,
+                    flatBufferBuilder.CreateVector(&operatorCode, 1),
+                    flatBufferBuilder.CreateVector(&subgraph, 1),
+                    modelDescription,
+                    flatBufferBuilder.CreateVector(buffers.data(), buffers.size()));
 
     flatBufferBuilder.Finish(flatbufferModel);
 
@@ -159,62 +114,46 @@ std::vector<char> CreateSliceTfLiteModel(tflite::TensorType tensorType,
 }
 
 template <typename T>
-void StridedSliceTestImpl(std::vector<armnn::BackendId>& backends,
-                          std::vector<T>& inputValues,
-                          std::vector<T>& expectedOutputValues,
-                          std::vector<int32_t>& beginTensorData,
-                          std::vector<int32_t>& endTensorData,
-                          std::vector<int32_t>& strideTensorData,
-                          std::vector<int32_t>& inputTensorShape,
-                          std::vector<int32_t>& beginTensorShape,
-                          std::vector<int32_t>& endTensorShape,
-                          std::vector<int32_t>& strideTensorShape,
-                          std::vector<int32_t>& outputTensorShape,
-                          const int32_t beginMask = 0,
-                          const int32_t endMask = 0,
-                          const int32_t ellipsisMask = 0,
-                          const int32_t newAxisMask = 0,
-                          const int32_t ShrinkAxisMask = 0,
-                          const armnn::DataLayout& dataLayout = armnn::DataLayout::NHWC)
+void SliceTestImpl(std::vector<armnn::BackendId>& backends,
+                   std::vector<T>& inputValues,
+                   std::vector<T>& expectedOutputValues,
+                   std::vector<int32_t>& beginTensorData,
+                   std::vector<int32_t>& sizeTensorData,
+                   std::vector<int32_t>& inputTensorShape,
+                   std::vector<int32_t>& beginTensorShape,
+                   std::vector<int32_t>& sizeTensorShape,
+                   std::vector<int32_t>& outputTensorShape)
 {
     using namespace tflite;
     std::vector<char> modelBuffer = CreateSliceTfLiteModel(
-            ::tflite::TensorType_FLOAT32,
-            inputTensorShape,
-            beginTensorData,
-            endTensorData,
-            strideTensorData,
-            beginTensorShape,
-            endTensorShape,
-            strideTensorShape,
-            outputTensorShape,
-            beginMask,
-            endMask,
-            ellipsisMask,
-            newAxisMask,
-            ShrinkAxisMask,
-            dataLayout);
+        ::tflite::TensorType_FLOAT32,
+        inputTensorShape,
+        beginTensorData,
+        sizeTensorData,
+        beginTensorShape,
+        sizeTensorShape,
+        outputTensorShape);
 
     auto tfLiteModel = GetModel(modelBuffer.data());
 
     // Create TfLite Interpreters
     std::unique_ptr<Interpreter> armnnDelegate;
     CHECK(InterpreterBuilder(tfLiteModel, ::tflite::ops::builtin::BuiltinOpResolver())
-              (&armnnDelegate) == kTfLiteOk);
+        (&armnnDelegate) == kTfLiteOk);
     CHECK(armnnDelegate != nullptr);
     CHECK(armnnDelegate->AllocateTensors() == kTfLiteOk);
 
     std::unique_ptr<Interpreter> tfLiteDelegate;
     CHECK(InterpreterBuilder(tfLiteModel, ::tflite::ops::builtin::BuiltinOpResolver())
-              (&tfLiteDelegate) == kTfLiteOk);
+        (&tfLiteDelegate) == kTfLiteOk);
     CHECK(tfLiteDelegate != nullptr);
     CHECK(tfLiteDelegate->AllocateTensors() == kTfLiteOk);
 
     // Create the ArmNN Delegate
     armnnDelegate::DelegateOptions delegateOptions(backends);
     std::unique_ptr<TfLiteDelegate, decltype(&armnnDelegate::TfLiteArmnnDelegateDelete)>
-    theArmnnDelegate(armnnDelegate::TfLiteArmnnDelegateCreate(delegateOptions),
-                     armnnDelegate::TfLiteArmnnDelegateDelete);
+        theArmnnDelegate(armnnDelegate::TfLiteArmnnDelegateCreate(delegateOptions),
+                         armnnDelegate::TfLiteArmnnDelegateDelete);
     CHECK(theArmnnDelegate != nullptr);
 
     // Modify armnnDelegateInterpreter to use armnnDelegate
@@ -236,6 +175,6 @@ void StridedSliceTestImpl(std::vector<armnn::BackendId>& backends,
 
     tfLiteDelegate.reset(nullptr);
     armnnDelegate.reset(nullptr);
-} // End of StridedSlice Test
+} // End of Slice Test
 
 } // anonymous namespace
\ No newline at end of file
diff --git a/delegate/src/test/StridedSliceTest.cpp b/delegate/src/test/StridedSliceTest.cpp
new file mode 100644
index 0000000000..43aea8a449
--- /dev/null
+++ b/delegate/src/test/StridedSliceTest.cpp
@@ -0,0 +1,241 @@
+//
+// Copyright © 2022 Arm Ltd and Contributors. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "StridedSliceTestHelper.hpp"
+
+#include <armnn_delegate.hpp>
+
+#include <flatbuffers/flatbuffers.h>
+
+#include <doctest/doctest.h>
+
+namespace armnnDelegate
+{
+
+void StridedSlice4DTest(std::vector<armnn::BackendId>& backends)
+{
+    std::vector<int32_t> inputShape  { 3, 2, 3, 1 };
+    std::vector<int32_t> outputShape { 1, 2, 3, 1 };
+    std::vector<int32_t> beginShape  { 4 };
+    std::vector<int32_t> endShape    { 4 };
+    std::vector<int32_t> strideShape { 4 };
+
+    std::vector<int32_t> beginData  { 1, 0, 0, 0 };
+    std::vector<int32_t> endData    { 2, 2, 3, 1 };
+    std::vector<int32_t> strideData { 1, 1, 1, 1 };
+    std::vector<float> inputData  { 1.0f, 1.0f, 1.0f, 2.0f, 2.0f, 2.0f,
+                                    3.0f, 3.0f, 3.0f, 4.0f, 4.0f, 4.0f,
+                                    5.0f, 5.0f, 5.0f, 6.0f, 6.0f, 6.0f };
+    std::vector<float> outputData { 3.0f, 3.0f, 3.0f, 4.0f, 4.0f, 4.0f };
+
+    StridedSliceTestImpl<float>(
+            backends,
+            inputData,
+            outputData,
+            beginData,
+            endData,
+            strideData,
+            inputShape,
+            beginShape,
+            endShape,
+            strideShape,
+            outputShape
+            );
+}
+
+void StridedSlice4DReverseTest(std::vector<armnn::BackendId>& backends)
+{
+    std::vector<int32_t> inputShape  { 3, 2, 3, 1 };
+    std::vector<int32_t> outputShape { 1, 2, 3, 1 };
+    std::vector<int32_t> beginShape  { 4 };
+    std::vector<int32_t> endShape    { 4 };
+    std::vector<int32_t> strideShape { 4 };
+
+    std::vector<int32_t> beginData  { 1, -1, 0, 0 };
+    std::vector<int32_t> endData    { 2, -3, 3, 1 };
+    std::vector<int32_t> strideData { 1, -1, 1, 1 };
+    std::vector<float>   inputData  { 1.0f, 1.0f, 1.0f, 2.0f, 2.0f, 2.0f,
+                                      3.0f, 3.0f, 3.0f, 4.0f, 4.0f, 4.0f,
+                                      5.0f, 5.0f, 5.0f, 6.0f, 6.0f, 6.0f };
+    std::vector<float>   outputData { 4.0f, 4.0f, 4.0f, 3.0f, 3.0f, 3.0f };
+
+    StridedSliceTestImpl<float>(
+            backends,
+            inputData,
+            outputData,
+            beginData,
+            endData,
+            strideData,
+            inputShape,
+            beginShape,
+            endShape,
+            strideShape,
+            outputShape
+    );
+}
+
+void StridedSliceSimpleStrideTest(std::vector<armnn::BackendId>& backends)
+{
+    std::vector<int32_t> inputShape  { 3, 2, 3, 1 };
+    std::vector<int32_t> outputShape { 2, 1, 2, 1 };
+    std::vector<int32_t> beginShape  { 4 };
+    std::vector<int32_t> endShape    { 4 };
+    std::vector<int32_t> strideShape { 4 };
+
+    std::vector<int32_t> beginData  { 0, 0, 0, 0 };
+    std::vector<int32_t> endData    { 3, 2, 3, 1 };
+    std::vector<int32_t> strideData { 2, 2, 2, 1 };
+    std::vector<float>   inputData  { 1.0f, 1.0f, 1.0f, 2.0f, 2.0f, 2.0f,
+                                      3.0f, 3.0f, 3.0f, 4.0f, 4.0f, 4.0f,
+                                      5.0f, 5.0f, 5.0f, 6.0f, 6.0f, 6.0f };
+    std::vector<float>   outputData { 1.0f, 1.0f,
+                                      5.0f, 5.0f };
+
+    StridedSliceTestImpl<float>(
+            backends,
+            inputData,
+            outputData,
+            beginData,
+            endData,
+            strideData,
+            inputShape,
+            beginShape,
+            endShape,
+            strideShape,
+            outputShape
+    );
+}
+
+void StridedSliceSimpleRangeMaskTest(std::vector<armnn::BackendId>& backends)
+{
+    std::vector<int32_t> inputShape  { 3, 2, 3, 1 };
+    std::vector<int32_t> outputShape { 3, 2, 3, 1 };
+    std::vector<int32_t> beginShape  { 4 };
+    std::vector<int32_t> endShape    { 4 };
+    std::vector<int32_t> strideShape { 4 };
+
+    std::vector<int32_t> beginData  { 1, 1, 1, 1 };
+    std::vector<int32_t> endData    { 1, 1, 1, 1 };
+    std::vector<int32_t> strideData { 1, 1, 1, 1 };
+
+    int beginMask = -1;
+    int endMask   = -1;
+
+    std::vector<float>   inputData  { 1.0f, 1.0f, 1.0f, 2.0f, 2.0f, 2.0f,
+                                      3.0f, 3.0f, 3.0f, 4.0f, 4.0f, 4.0f,
+                                      5.0f, 5.0f, 5.0f, 6.0f, 6.0f, 6.0f };
+    std::vector<float>   outputData { 1.0f, 1.0f, 1.0f, 2.0f, 2.0f, 2.0f,
+                                      3.0f, 3.0f, 3.0f, 4.0f, 4.0f, 4.0f,
+                                      5.0f, 5.0f, 5.0f, 6.0f, 6.0f, 6.0f };
+
+    StridedSliceTestImpl<float>(
+            backends,
+            inputData,
+            outputData,
+            beginData,
+            endData,
+            strideData,
+            inputShape,
+            beginShape,
+            endShape,
+            strideShape,
+            outputShape,
+            beginMask,
+            endMask
+    );
+}
+
+TEST_SUITE("StridedSlice_CpuRefTests")
+{
+
+TEST_CASE ("StridedSlice_4D_CpuRef_Test")
+{
+    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuRef};
+    StridedSlice4DTest(backends);
+}
+
+TEST_CASE ("StridedSlice_4D_Reverse_CpuRef_Test")
+{
+    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuRef};
+    StridedSlice4DReverseTest(backends);
+}
+
+TEST_CASE ("StridedSlice_SimpleStride_CpuRef_Test")
+{
+    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuRef};
+    StridedSliceSimpleStrideTest(backends);
+}
+
+TEST_CASE ("StridedSlice_SimpleRange_CpuRef_Test")
+{
+    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuRef};
+    StridedSliceSimpleRangeMaskTest(backends);
+}
+
+} // StridedSlice_CpuRefTests TestSuite
+
+
+
+TEST_SUITE("StridedSlice_CpuAccTests")
+{
+
+TEST_CASE ("StridedSlice_4D_CpuAcc_Test")
+{
+    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuAcc};
+    StridedSlice4DTest(backends);
+}
+
+TEST_CASE ("StridedSlice_4D_Reverse_CpuAcc_Test")
+{
+    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuAcc};
+    StridedSlice4DReverseTest(backends);
+}
+
+TEST_CASE ("StridedSlice_SimpleStride_CpuAcc_Test")
+{
+    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuAcc};
+    StridedSliceSimpleStrideTest(backends);
+}
+
+TEST_CASE ("StridedSlice_SimpleRange_CpuAcc_Test")
+{
+    std::vector<armnn::BackendId> backends = {armnn::Compute::CpuAcc};
+    StridedSliceSimpleRangeMaskTest(backends);
+}
+
+} // StridedSlice_CpuAccTests TestSuite
+
+
+
+TEST_SUITE("StridedSlice_GpuAccTests")
+{
+
+TEST_CASE ("StridedSlice_4D_GpuAcc_Test")
+{
+    std::vector<armnn::BackendId> backends = {armnn::Compute::GpuAcc};
+    StridedSlice4DTest(backends);
+}
+
+TEST_CASE ("StridedSlice_4D_Reverse_GpuAcc_Test")
+{
+    std::vector<armnn::BackendId> backends = {armnn::Compute::GpuAcc};
+    StridedSlice4DReverseTest(backends);
+}
+
+TEST_CASE ("StridedSlice_SimpleStride_GpuAcc_Test")
+{
+    std::vector<armnn::BackendId> backends = {armnn::Compute::GpuAcc};
+    StridedSliceSimpleStrideTest(backends);
+}
+
+TEST_CASE ("StridedSlice_SimpleRange_GpuAcc_Test")
+{
+    std::vector<armnn::BackendId> backends = {armnn::Compute::GpuAcc};
+    StridedSliceSimpleRangeMaskTest(backends);
+}
+
+} // StridedSlice_GpuAccTests TestSuite
+
+} // namespace armnnDelegate
\ No newline at end of file
diff --git a/delegate/src/test/StridedSliceTestHelper.hpp b/delegate/src/test/StridedSliceTestHelper.hpp
new file mode 100644
index 0000000000..2bca4fdc35
--- /dev/null
+++ b/delegate/src/test/StridedSliceTestHelper.hpp
@@ -0,0 +1,218 @@
+//
+// Copyright © 2022 Arm Ltd and Contributors. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include "TestUtils.hpp"
+
+#include <armnn_delegate.hpp>
+#include <armnn/DescriptorsFwd.hpp>
+
+#include <flatbuffers/flatbuffers.h>
+#include <tensorflow/lite/interpreter.h>
+#include <tensorflow/lite/kernels/register.h>
+#include <tensorflow/lite/model.h>
+#include <tensorflow/lite/schema/schema_generated.h>
+#include <tensorflow/lite/version.h>
+
+#include <doctest/doctest.h>
+
+#include <string>
+
+namespace
+{
+
+std::vector<char> CreateStridedSliceTfLiteModel(tflite::TensorType tensorType,
+                                                const std::vector<int32_t>& inputTensorShape,
+                                                const std::vector<int32_t>& beginTensorData,
+                                                const std::vector<int32_t>& endTensorData,
+                                                const std::vector<int32_t>& strideTensorData,
+                                                const std::vector<int32_t>& beginTensorShape,
+                                                const std::vector<int32_t>& endTensorShape,
+                                                const std::vector<int32_t>& strideTensorShape,
+                                                const std::vector<int32_t>& outputTensorShape,
+                                                const int32_t beginMask,
+                                                const int32_t endMask,
+                                                const int32_t ellipsisMask,
+                                                const int32_t newAxisMask,
+                                                const int32_t ShrinkAxisMask,
+                                                const armnn::DataLayout& dataLayout)
+{
+    using namespace tflite;
+    flatbuffers::FlatBufferBuilder flatBufferBuilder;
+
+    std::array<flatbuffers::Offset<tflite::Buffer>, 4> buffers;
+    buffers[0] = CreateBuffer(flatBufferBuilder, flatBufferBuilder.CreateVector({}));
+    buffers[1] = CreateBuffer(flatBufferBuilder,
+                              flatBufferBuilder.CreateVector(reinterpret_cast<const uint8_t*>(beginTensorData.data()),
+                                                             sizeof(int32_t) * beginTensorData.size()));
+    buffers[2] = CreateBuffer(flatBufferBuilder,
+                              flatBufferBuilder.CreateVector(reinterpret_cast<const uint8_t*>(endTensorData.data()),
+                                                             sizeof(int32_t) * endTensorData.size()));
+    buffers[3] = CreateBuffer(flatBufferBuilder,
+                              flatBufferBuilder.CreateVector(reinterpret_cast<const uint8_t*>(strideTensorData.data()),
+                                                             sizeof(int32_t) * strideTensorData.size()));
+
+    std::array<flatbuffers::Offset<Tensor>, 5> tensors;
+    tensors[0] = CreateTensor(flatBufferBuilder,
+                              flatBufferBuilder.CreateVector<int32_t>(inputTensorShape.data(),
+                                                                      inputTensorShape.size()),
+                              tensorType,
+                              0,
+                              flatBufferBuilder.CreateString("input"));
+    tensors[1] = CreateTensor(flatBufferBuilder,
+                              flatBufferBuilder.CreateVector<int32_t>(beginTensorShape.data(),
+                                                                      beginTensorShape.size()),
+                              ::tflite::TensorType_INT32,
+                              1,
+                              flatBufferBuilder.CreateString("begin_tensor"));
+    tensors[2] = CreateTensor(flatBufferBuilder,
+                              flatBufferBuilder.CreateVector<int32_t>(endTensorShape.data(),
+                                                                      endTensorShape.size()),
+                              ::tflite::TensorType_INT32,
+                              2,
+                              flatBufferBuilder.CreateString("end_tensor"));
+    tensors[3] = CreateTensor(flatBufferBuilder,
+                              flatBufferBuilder.CreateVector<int32_t>(strideTensorShape.data(),
+                                                                      strideTensorShape.size()),
+                              ::tflite::TensorType_INT32,
+                              3,
+                              flatBufferBuilder.CreateString("stride_tensor"));
+    tensors[4] = CreateTensor(flatBufferBuilder,
+                              flatBufferBuilder.CreateVector<int32_t>(outputTensorShape.data(),
+                                                                      outputTensorShape.size()),
+                              tensorType,
+                              0,
+                              flatBufferBuilder.CreateString("output"));
+
+
+    // create operator
+    tflite::BuiltinOptions operatorBuiltinOptionsType = tflite::BuiltinOptions_StridedSliceOptions;
+    flatbuffers::Offset<void> operatorBuiltinOptions = CreateStridedSliceOptions(flatBufferBuilder,
+                                                                                 beginMask,
+                                                                                 endMask,
+                                                                                 ellipsisMask,
+                                                                                 newAxisMask,
+                                                                                 ShrinkAxisMask).Union();
+
+    const std::vector<int> operatorInputs{ 0, 1, 2, 3 };
+    const std::vector<int> operatorOutputs{ 4 };
+    flatbuffers::Offset <Operator> sliceOperator =
+            CreateOperator(flatBufferBuilder,
+                           0,
+                           flatBufferBuilder.CreateVector<int32_t>(operatorInputs.data(), operatorInputs.size()),
+                           flatBufferBuilder.CreateVector<int32_t>(operatorOutputs.data(), operatorOutputs.size()),
+                           operatorBuiltinOptionsType,
+                           operatorBuiltinOptions);
+
+    const std::vector<int> subgraphInputs{ 0, 1, 2, 3 };
+    const std::vector<int> subgraphOutputs{ 4 };
+    flatbuffers::Offset <SubGraph> subgraph =
+            CreateSubGraph(flatBufferBuilder,
+                           flatBufferBuilder.CreateVector(tensors.data(), tensors.size()),
+                           flatBufferBuilder.CreateVector<int32_t>(subgraphInputs.data(), subgraphInputs.size()),
+                           flatBufferBuilder.CreateVector<int32_t>(subgraphOutputs.data(), subgraphOutputs.size()),
+                           flatBufferBuilder.CreateVector(&sliceOperator, 1));
+
+    flatbuffers::Offset <flatbuffers::String> modelDescription =
+            flatBufferBuilder.CreateString("ArmnnDelegate: StridedSlice Operator Model");
+    flatbuffers::Offset <OperatorCode> operatorCode = CreateOperatorCode(flatBufferBuilder,
+                                                                         BuiltinOperator_STRIDED_SLICE);
+
+    flatbuffers::Offset <Model> flatbufferModel =
+            CreateModel(flatBufferBuilder,
+                        TFLITE_SCHEMA_VERSION,
+                        flatBufferBuilder.CreateVector(&operatorCode, 1),
+                        flatBufferBuilder.CreateVector(&subgraph, 1),
+                        modelDescription,
+                        flatBufferBuilder.CreateVector(buffers.data(), buffers.size()));
+
+    flatBufferBuilder.Finish(flatbufferModel);
+
+    return std::vector<char>(flatBufferBuilder.GetBufferPointer(),
+                             flatBufferBuilder.GetBufferPointer() + flatBufferBuilder.GetSize());
+}
+
+template <typename T>
+void StridedSliceTestImpl(std::vector<armnn::BackendId>& backends,
+                          std::vector<T>& inputValues,
+                          std::vector<T>& expectedOutputValues,
+                          std::vector<int32_t>& beginTensorData,
+                          std::vector<int32_t>& endTensorData,
+                          std::vector<int32_t>& strideTensorData,
+                          std::vector<int32_t>& inputTensorShape,
+                          std::vector<int32_t>& beginTensorShape,
+                          std::vector<int32_t>& endTensorShape,
+                          std::vector<int32_t>& strideTensorShape,
+                          std::vector<int32_t>& outputTensorShape,
+                          const int32_t beginMask = 0,
+                          const int32_t endMask = 0,
+                          const int32_t ellipsisMask = 0,
+                          const int32_t newAxisMask = 0,
+                          const int32_t ShrinkAxisMask = 0,
+                          const armnn::DataLayout& dataLayout = armnn::DataLayout::NHWC)
+{
+    using namespace tflite;
+    std::vector<char> modelBuffer = CreateStridedSliceTfLiteModel(
+            ::tflite::TensorType_FLOAT32,
+            inputTensorShape,
+            beginTensorData,
+            endTensorData,
+            strideTensorData,
+            beginTensorShape,
+            endTensorShape,
+            strideTensorShape,
+            outputTensorShape,
+            beginMask,
+            endMask,
+            ellipsisMask,
+            newAxisMask,
+            ShrinkAxisMask,
+            dataLayout);
+
+    auto tfLiteModel = GetModel(modelBuffer.data());
+
+    // Create TfLite Interpreters
+    std::unique_ptr<Interpreter> armnnDelegate;
+    CHECK(InterpreterBuilder(tfLiteModel, ::tflite::ops::builtin::BuiltinOpResolver())
+              (&armnnDelegate) == kTfLiteOk);
+    CHECK(armnnDelegate != nullptr);
+    CHECK(armnnDelegate->AllocateTensors() == kTfLiteOk);
+
+    std::unique_ptr<Interpreter> tfLiteDelegate;
+    CHECK(InterpreterBuilder(tfLiteModel, ::tflite::ops::builtin::BuiltinOpResolver())
+              (&tfLiteDelegate) == kTfLiteOk);
+    CHECK(tfLiteDelegate != nullptr);
+    CHECK(tfLiteDelegate->AllocateTensors() == kTfLiteOk);
+
+    // Create the ArmNN Delegate
+    armnnDelegate::DelegateOptions delegateOptions(backends);
+    std::unique_ptr<TfLiteDelegate, decltype(&armnnDelegate::TfLiteArmnnDelegateDelete)>
+                                   theArmnnDelegate(armnnDelegate::TfLiteArmnnDelegateCreate(delegateOptions),
+                                                    armnnDelegate::TfLiteArmnnDelegateDelete);
+    CHECK(theArmnnDelegate != nullptr);
+
+    // Modify armnnDelegateInterpreter to use armnnDelegate
+    CHECK(armnnDelegate->ModifyGraphWithDelegate(theArmnnDelegate.get()) == kTfLiteOk);
+
+    // Set input data
+    armnnDelegate::FillInput<T>(tfLiteDelegate, 0, inputValues);
+    armnnDelegate::FillInput<T>(armnnDelegate, 0, inputValues);
+
+    // Run EnqueWorkload
+    CHECK(tfLiteDelegate->Invoke() == kTfLiteOk);
+    CHECK(armnnDelegate->Invoke() == kTfLiteOk);
+
+    // Compare output data
+    armnnDelegate::CompareOutputData<T>(tfLiteDelegate,
+                                        armnnDelegate,
+                                        outputTensorShape,
+                                        expectedOutputValues);
+
+    tfLiteDelegate.reset(nullptr);
+    armnnDelegate.reset(nullptr);
+} // End of StridedSlice Test
+
+} // anonymous namespace
\ No newline at end of file
diff --git a/src/armnnTfLiteParser/test/Slice.cpp b/src/armnnTfLiteParser/test/Slice.cpp
index a2a791feef..83c0b73d2f 100644
--- a/src/armnnTfLiteParser/test/Slice.cpp
+++ b/src/armnnTfLiteParser/test/Slice.cpp
@@ -174,9 +174,9 @@ TEST_CASE_FIXTURE(SliceFixtureD213, "SliceD213")
 struct DynamicSliceFixtureD213 : SliceFixture
 {
     DynamicSliceFixtureD213() : SliceFixture("[ 3, 2, 3 ]",
-                                            "[ ]",
-                                              "[ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]",
-                                                "[ 255, 255, 255, 255, 1, 0, 0, 0, 255, 255, 255, 255 ]") {}
+                                             "[ ]",
+                                             "[ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]",
+                                             "[ 255, 255, 255, 255, 1, 0, 0, 0, 255, 255, 255, 255 ]") {}
 };
 
 TEST_CASE_FIXTURE(DynamicSliceFixtureD213, "DynamicSliceD213")