MIVisionX ONNX Model Validation#


    MIVisionX-WinML-Validate.exe [options]  --m <ONNX.model full path>
                                            --i <model input tensor name>
                                            --o <model output tensor name>
                                            --s <output tensor size in (n,c,h,w)>
                                            --l <label.txt full path>
                                            --f <image frame full path>
                                            --d <Learning Model Device Kind <DirectXHighPerformance>> [optional]

MIVisionX ONNX Model Validation Parameters#

    --m/--model                     -- onnx model full path [required]
    --i/--inputName                 -- model input tensor name [required]
    --o/--outputName                -- model output tensor name [required]
    --s/--outputSize                -- model output tensor size <n,c,h,w> [required]
    --l/--label                     -- label.txt file full path [required]
    --f/--imageFrame                -- imageFrame.png file full path [required]
    --d/--deviceKind                -- Learning Model Device Kind <0-4> [optional]
                                     0 - Default
                                     1 - Cpu
                                     2 - DirectX
                                     3 - DirectXHighPerformance
                                     4 - DirectXMinPower

MIVisionX ONNX Model Validation Options#

    --h/--help      -- Show full help


Get ONNX models from ONNX Model Zoo

Sample - SqeezeNet#

  • Download the SqueezeNet ONNX Model

  • Use Netron to open the model.onnx

    • Look at Model Properties to find Input & Output Tensor Name (data_0 - input; softmaxout_1 - output)

    • Look at output tensor dimensions (n,c,h,w - [1,1000,1,1] for softmaxout_1)

  • Use the label file - Labels.txt and sample image - car.JPEG to run the MIVisionX WinML Validation

  • Use –d 0 if only CPU available, else use –d 3 for GPU inference

        MIVisionX-WinML-Validate.exe [options]  --m \full-path-to-model\model.onnx
                                                --i data_0
                                                --o softmaxout_1
                                                --s 1,1000,1,1
                                                --l \full-path-to-labels\Labels.txt 
                                                --f \full-path-to-labels\car.JPEG
                                                --d 3 [optional]