Fix preprocessing

Author: nums11

Samples/WinMLSamplesGallery/WinMLSamplesGalleryNative/D3D12Quad.cpp

@@ -157,13 +157,13 @@ void D3D12Quad::LoadAssets()
     m_commandList->ResourceBarrier(1, &present_to_copy_src);
 
     auto desc = m_renderTargets[m_frameIndex]->GetDesc();
-    UINT64 rowSize, totalSize;
-    m_device->GetCopyableFootprints(&desc, 0, 1, 0, nullptr, nullptr, &rowSize, &totalSize);
+    UINT64 rowSizeInBytes, totalSizeInBytes;
+    m_device->GetCopyableFootprints(&desc, 0, 1, 0, nullptr, nullptr, &rowSizeInBytes, &totalSizeInBytes);
     D3D12_PLACED_SUBRESOURCE_FOOTPRINT bufferFootprint = {};
     bufferFootprint.Footprint.Width = static_cast<UINT>(desc.Width);
     bufferFootprint.Footprint.Height = desc.Height;
     bufferFootprint.Footprint.Depth = 1;
-    bufferFootprint.Footprint.RowPitch = static_cast<UINT>((rowSize + 255) & ~255);
+    bufferFootprint.Footprint.RowPitch = static_cast<UINT>((rowSizeInBytes + 255) & ~255);
     bufferFootprint.Footprint.Format = desc.Format;
 
     const CD3DX12_TEXTURE_COPY_LOCATION copyDest(currentBuffer.Get(), bufferFootprint);
@@ -395,19 +395,19 @@ void D3D12Quad::CreateCurrentBuffer()
     D3D12_RESOURCE_DESC bufferDesc = {};
     bufferDesc.DepthOrArraySize = 1;
     bufferDesc.Dimension = D3D12_RESOURCE_DIMENSION_BUFFER;
-    bufferDesc.Flags = D3D12_RESOURCE_FLAG_NONE;
+    bufferDesc.Flags = D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS;
     bufferDesc.Format = DXGI_FORMAT_UNKNOWN;
     //bufferDesc.Format = DXGI_FORMAT_R8G8B8A8_UNORM;
     bufferDesc.Height = 1;
-    bufferDesc.Width = ((desc.Width + 255) & ~255) * 4 * desc.Height;
+    bufferDesc.Width = ((desc.Width* 4  + 255) & ~255) * desc.Height;
     //bufferDesc.Width = static_cast<uint64_t>(800 * 600 * 3 * 4);
     bufferDesc.Layout = D3D12_TEXTURE_LAYOUT_ROW_MAJOR;
     //bufferDesc.Layout = D3D12_TEXTURE_LAYOUT_UNKNOWN;
     bufferDesc.MipLevels = 1;
     bufferDesc.SampleDesc.Count = 1;
 
     //auto heap_properties = CD3DX12_HEAP_PROPERTIES(D3D12_HEAP_TYPE_DEFAULT);
-    const CD3DX12_HEAP_PROPERTIES readBackHeapProperties(D3D12_HEAP_TYPE_READBACK);
+    const CD3DX12_HEAP_PROPERTIES readBackHeapProperties(D3D12_HEAP_TYPE_DEFAULT);
     auto some_hr = m_device->CreateCommittedResource(
         &readBackHeapProperties,
         D3D12_HEAP_FLAG_NONE,
@@ -535,13 +535,13 @@ void D3D12Quad::PopulateCommandList()
     //m_commandList->CopyResource(currentBuffer.Get(), m_renderTargets[m_frameIndex].Get());
 
     auto desc = m_renderTargets[m_frameIndex]->GetDesc();
-    UINT64 rowSize, totalSize;
-    m_device->GetCopyableFootprints(&desc, 0, 1, 0, nullptr, nullptr, &rowSize, &totalSize);
+    UINT64 rowSizeInBytes, totalSizeInBytes;
+    m_device->GetCopyableFootprints(&desc, 0, 1, 0, nullptr, nullptr, &rowSizeInBytes, &totalSizeInBytes);
     D3D12_PLACED_SUBRESOURCE_FOOTPRINT bufferFootprint = {};
     bufferFootprint.Footprint.Width = desc.Width;
     bufferFootprint.Footprint.Height = desc.Height;
     bufferFootprint.Footprint.Depth = 1;
-    bufferFootprint.Footprint.RowPitch = static_cast<UINT>((rowSize + 255) & ~255);
+    bufferFootprint.Footprint.RowPitch = static_cast<UINT>((rowSizeInBytes + 255) & ~255);
     bufferFootprint.Footprint.Format = desc.Format;
 
     const CD3DX12_TEXTURE_COPY_LOCATION copyDest(currentBuffer.Get(), bufferFootprint);

Samples/WinMLSamplesGallery/WinMLSamplesGalleryNative/DXResourceBinding.cpp

@@ -23,7 +23,7 @@ using TensorKind = winrt::Microsoft::AI::MachineLearning::TensorKind;
 using LearningModelBuilder = winrt::Microsoft::AI::MachineLearning::Experimental::LearningModelBuilder;
 using LearningModelOperator = winrt::Microsoft::AI::MachineLearning::Experimental::LearningModelOperator;
 
-std::array<int64_t, 4> preprocessInputShape;
+std::array<int64_t, 2> preprocessInputShape;
 
 std::array<long, 6> CalculateCenterFillDimensions(long oldH, long oldW, long h, long w)
 {
@@ -81,12 +81,12 @@ namespace winrt::WinMLSamplesGalleryNative::implementation
         long left = center_fill_dimensions[4];
         long right = center_fill_dimensions[5];
         winrt::hstring interpolationMode = L"nearest";
-        long c = 3;
+        long c = 4;
 
         auto width = 800;
         auto height = 600;
-        auto rowPitchInPixels = (width + 255) & ~255;
-        auto rowPitchInBytes = rowPitchInPixels * 4;
+        auto rowPitchInBytes = (width * 4 + 255) & ~255;
+        auto rowPitchInPixels = rowPitchInBytes / 4;
         auto bufferInBytes = rowPitchInBytes * height;
         preprocessInputShape = { 1, bufferInBytes};
         //const std::array<int64_t, 4> preprocessInputShape = { 1, 512, 512, 4 };
@@ -158,7 +158,7 @@ namespace winrt::WinMLSamplesGalleryNative::implementation
 
         auto reshape_op = LearningModelOperator(L"Reshape")
             .SetInput(L"data", L"CastOutput")
-            .SetConstant(L"shape", TensorInt64Bit::CreateFromIterable({ 4 }, { 1, 512, 512, 4 }))
+            .SetConstant(L"shape", TensorInt64Bit::CreateFromIterable({ 4 }, { 1, height, 832, 4 }))
             .SetOutput(L"reshaped", L"ReshapeOutput");
 
         auto slice_1 = LearningModelOperator(L"Slice")
@@ -182,7 +182,7 @@ namespace winrt::WinMLSamplesGalleryNative::implementation
 
         auto preprocessingModelBuilder =
             LearningModelBuilder::Create(12)
-            .Inputs().Add(LearningModelBuilder::CreateTensorFeatureDescriptor(L"Input", TensorKind::Float, preprocessInputShape))
+            .Inputs().Add(LearningModelBuilder::CreateTensorFeatureDescriptor(L"Input", TensorKind::UInt8, preprocessInputShape))
             .Outputs().Add(LearningModelBuilder::CreateTensorFeatureDescriptor(L"Output", TensorKind::Float, preprocessOutputShape))
             .Operators().Add(cast_op)
             .Operators().Add(reshape_op)

Samples/WinMLSamplesGallery/WinMLSamplesGalleryNative/ORTHelpers.cpp

@@ -48,7 +48,7 @@ Ort::Session CreateSession(const wchar_t* model_file_path)
 // image from 512 x 512 x 4 to 224 x 224 x 3
 Ort::Value Preprocess(Ort::Session& session,
     ComPtr<ID3D12Resource> currentBuffer,
-    const std::array<int64_t, 4> inputShape)
+    const std::array<int64_t, 2> inputShape)
 {
     // Init OrtAPI
     OrtApi const& ortApi = Ort::GetApi(); // Uses ORT_API_VERSION
@@ -65,7 +65,7 @@ Ort::Value Preprocess(Ort::Session& session,
         memoryInformation,
         currentBuffer.Get(),
         inputShape,
-        ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT,
+        ONNX_TENSOR_ELEMENT_DATA_TYPE_UINT8,
         /*out*/ IID_PPV_ARGS_Helper(inputTensorEpWrapper.GetAddressOf())
     );
 

Samples/WinMLSamplesGallery/WinMLSamplesGalleryNative/ORTHelpers.h

@@ -14,7 +14,7 @@ Ort::Session CreateSession(const wchar_t* model_file_path);
 
 Ort::Value Preprocess(Ort::Session& session,
     ComPtr<ID3D12Resource> currentBuffer,
-    const std::array<int64_t, 4> inputShape);
+    const std::array<int64_t, 2> inputShape);
 
 winrt::com_array<float> Eval(Ort::Session& session, const Ort::Value& prev_input);