SportDX Docker Container Setup

This document provides instructions for building and running the sportdx Docker container, which is set up for GPU-accelerated PyTorch applications with GUI support via WSL2.

Prerequisites

• Windows 10 or 11 with WSL2 enabled.
• NVIDIA GPU with the latest drivers installed.
• Docker Desktop for Windows with WSL2 backend and NVIDIA Container Toolkit.

Building the Docker Image

1. Open WSL2 and navigate to the directory containing the Dockerfile (after installing and running docker).
2. Build the Docker image:

   docker build -t sportdx .

Running the Docker Container

To run the sportdx container with GPU and GUI support:

1. Allow local connections to the X server:

   xhost +local:docker

2. Run the container: (only use --gpus all if you have a gpu)

   docker run -it --gpus all -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix -v $(pwd):/usr/src/app/KI-basierte-Erkennung-von-Bewegungen sportdx

docker run -it --gpus all -v $(pwd):/usr/src/app/KI-basierte-Erkennung-von-Bewegungen sportdx

Starting an Existing Container

If the container sportdx is already created and you want to start it:

1. Start the container:

   docker start sportdx

2. Attach to the container for interaction:

   docker attach sportdx

Folder Structure

├── checkpoints *Place motionbert checkpoint here*
├── configs *place config motionBert here*
├── models *yolov8 model will be here*
├── MotionBERT4sportDX *Motionbert fork*
├── outputs *outputs from yolov8 and motionbert*
├── videos/todo *add videos here*
├── dockerfile
├── main.py
├── README.md
└── requirements.txt

Notes

• The command xhost +local:docker opens up the X server for local connections and should be used with caution due to potential security implications.
• The Docker setup is advanced and might require specific configurations based on your hardware and software environment.
• Ensure that your WSL2 and Docker Desktop are properly configured for GPU acceleration.