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.

1. Environment

1. Docker 20.10.5, build 55c4c88

2. Model optimization environment to be built

1. Ubuntu 18.04 x86_64
2. TensorFlow v2.4.1 (MediaPipe Custom OP, FlexDelegate, XNNPACK enabled)
3. tflite_runtime v2.4.1 (MediaPipe Custom OP, FlexDelegate, XNNPACK enabled)
4. edgetpu-compiler
5. flatc 1.12.0
6. TensorRT cuda11.0-trt7.1.3.4-ga-20200617
7. PyTorch 1.7.1+cu110
8. TorchVision 0.8.2+cu110
9. TorchAudio 0.7.2
10. OpenVINO 2021.3.394
11. tensorflowjs
12. coremltools
13. onnx
14. tf2onnx
15. tensorflow-datasets
16. openvino2tensorflow
17. tflite2tensorflow
18. onnxruntime
19. onnx-simplifier
20. gdown
21. OpenCV 4.5.2-openvino

3. Usage

3-1. Docker Hub

$ 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 \
    -e DISPLAY=$DISPLAY \
    --privileged \
    pinto0309/mtomo:ubuntu1804_tf2.4.1_torch1.7.1_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 \
    -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