mtomo

Multiple types of NN model optimization environments. It is possible to directly access the host PC GUI and the camera to verify the operation. And, Intel iHD GPU (iGPU) support. NVIDIA GPU (dGPU) support.

1. Environment

1. Docker 20.10.5, build 55c4c88

2. Model optimization environment to be built

1. Ubuntu 20.04 x86_64
2. CUDA 11.2
3. cuDNN 8.1
4. TensorFlow v2.5.0-rc1 (MediaPipe Custom OP, FlexDelegate, XNNPACK enabled)
5. tflite_runtime v2.5.0-rc1 (MediaPipe Custom OP, FlexDelegate, XNNPACK enabled)
6. edgetpu-compiler
7. flatc 1.12.0
8. TensorRT cuda11.1-trt7.2.3.4-ga-20210226
9. PyTorch 1.8.1+cu112
10. TorchVision 0.9.1+cu112
11. TorchAudio 0.8.1
12. OpenVINO 2021.3.394
13. tensorflowjs
14. coremltools
15. onnx
16. tf2onnx
17. tensorflow-datasets
18. openvino2tensorflow
19. tflite2tensorflow
20. onnxruntime
21. onnx-simplifier
22. MXNet
23. gdown
24. OpenCV 4.5.2-openvino
25. Intel-Media-SDK
26. Intel iHD GPU (iGPU) support

3. Usage

3-1. Docker Hub

https://hub.docker.com/repository/docker/pinto0309/mtomo/tags?page=1&ordering=last_updated

$ xhost +local: && \
  docker run -it --rm \
    --gpus all \
    -v `pwd`:/home/user/workdir \
    -v /tmp/.X11-unix/:/tmp/.X11-unix:rw \
    --device /dev/video0:/dev/video0:mwr \
    --net=host \
    -e LIBVA_DRIVER_NAME=iHD \
    -e XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR \
    -e DISPLAY=$DISPLAY \
    --privileged \
    pinto0309/mtomo:ubuntu2004_tf2.5.0-rc1_torch1.8.1_openvino2021.3.394

3-2. Docker Build

$ git clone https://github.com/PINTO0309/mtomo.git && cd mtomo
$ docker build -t {IMAGE_NAME}:{TAG} .

3-3. Docker Run

$ xhost +local: && \
  docker run -it --rm \
    --gpus all \
    -v `pwd`:/home/user/workdir \
    -v /tmp/.X11-unix/:/tmp/.X11-unix:rw \
    --device /dev/video0:/dev/video0:mwr \
    --net=host \
    -e LIBVA_DRIVER_NAME=iHD \
    -e XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR \
    -e DISPLAY=$DISPLAY \
    --privileged \
    {IMAGE_NAME}:{TAG}

4. Reference articles

1. openvino2tensorflow
2. tflite2tensorflow
3. tensorflow-onnx (a.k.a tf2onnx)
4. tensorflowjs
5. coremltools
6. OpenVINO
7. onnx
8. onnx-simplifier
9. TensorFLow
10. PyTorch
11. flatbuffers (a.k.a flatc)
12. TensorRT
13. Intel-Media-SDK/MediaSDK - Running on GPU under docker
14. Intel-Media-SDK/MediaSDK - Intel media stack on Ubuntu