RapidPoseTriangulation/extras/jetson/README.md

Setup with Nvidia-Jetson-Orin

Initial setup and installation of RapidPoseTriangulation on a Nvidia Jetson device.
Tested with a Jetson AGX Orin Developer Kit module.

Base installation

• Install newest software image:
  (https://developer.nvidia.com/sdk-manager)

• Initialize system:
  (https://developer.nvidia.com/embedded/learn/get-started-jetson-agx-orin-devkit)

• Install basic tools:

  sudo apt install -y curl nano wget git
  sudo apt install -y terminator
  

• Test docker is working:

  sudo docker run --rm hello-world
  

• Enable docker without sudo:
  (https://docs.docker.com/engine/install/linux-postinstall/#manage-docker-as-a-non-root-user)

• Enable GPU-access for docker building:

  Run sudo nano /etc/docker/daemon.json and add:

  {
    "runtimes": {
      "nvidia": {
        "args": [],
        "path": "nvidia-container-runtime"
      }
    },
    "default-runtime": "nvidia"
  }
  

  Restart docker: sudo systemctl restart docker

• Install vs-code:
  (https://code.visualstudio.com/docs/setup/linux)

• Test docker is working:

  docker run --rm hello-world
  docker run -it --rm --runtime=nvidia --network=host -e DISPLAY=$DISPLAY -v /tmp/.X11-unix/:/tmp/.X11-unix nvcr.io/nvidia/l4t-base:r36.2.0
  docker run -it --rm --runtime=nvidia --network=host dustynv/onnxruntime:1.20-r36.4.0
  

• Check cuda access in container:

  python3 -c 'import torch; print(torch.cuda.is_available());'
  

• Enable maximum performance mode:

  sudo nvpmodel -m 0
  sudo jetson_clocks
  

RPT installation

• Build docker container:

  docker build --progress=plain -f extras/jetson/dockerfile -t rapidposetriangulation .
  ./run_container.sh
  

• Build rpt package inside container:

  cd /RapidPoseTriangulation/swig/ && make all && cd ../tests/ && python3 test_interface.py && cd ..
  
  cd /RapidPoseTriangulation/scripts/ && \
  g++ -std=c++2a -fPIC -O3 -march=native -Wall -Werror -flto=auto -fopenmp -fopenmp-simd \
    -I /RapidPoseTriangulation/rpt/ \
    -isystem /usr/include/opencv4/ \
    -isystem /usr/local/include/onnxruntime/ \
    -L /usr/local/lib/ \
    test_skelda_dataset.cpp \
    /RapidPoseTriangulation/rpt/*.cpp \
    -o test_skelda_dataset.bin \
    -Wl,--start-group \
      -lonnxruntime_providers_tensorrt \
      -lonnxruntime_providers_shared \
      -lonnxruntime_providers_cuda \
      -lonnxruntime \
    -Wl,--end-group \
    $(pkg-config --libs opencv4) \
    -Wl,-rpath,/onnxruntime/build/Linux/Release/ \
  && cd ..
  

• Test with samples:

  python3 /RapidPoseTriangulation/scripts/test_triangulate.py
  

• Evaluate datasets:

  python3 /RapidPoseTriangulation/scripts/test_skelda_dataset.py
  

ROS interface

• Build docker container:

  docker build --progress=plain -f extras/jetson/dockerfile_ros -t rapidposetriangulation_ros .
  ./run_container.sh
  

• Run and test:

  xhost +; docker compose -f extras/jetson/docker-compose.yml up
  
  docker exec -it ros-test_node-1 bash
  export ROS_DOMAIN_ID=18