This project uses MediaPipe and OpenCV to detect and visualize facial landmarks in video frames. It leverages MediaPipe's FaceMesh model to track keypoints and overlays them on the input video.
- Real-time face mesh detection on video input
- Visual overlay of facial landmarks using OpenCV
- Dockerized setup for easy deployment
- Qt-based GUI rendering inside container (with X11 support)
- Docker installed on your system
- X11 server running (for GUI display)
(Note: Extra GUI steps are not needed if you are running this on windows locally, but this was developed in WSL.)
docker build -t face-keypoint-tracking .
2. Allow Docker to access your display
On Linux:
xhost +local:docker
On WSL (Windows Subsystem for Linux):
export DISPLAY=$(cat /etc/resolv.conf | grep nameserver | awk '{print $2}'):0
xhost +local:docker
3. Run the container
docker run -e DISPLAY=$DISPLAY \
-v /tmp/.X11-unix:/tmp/.X11-unix \
--net=host \
face-keypoint-tracking
face-keypoint-detection/
├── Dockerfile
├── requirements.txt
├── main.py
└── muhammad_ali.mkv # Replace with your own video file
Once the container starts, it will:
- Load the video file muhammad_ali.mkv
- Detect facial landmarks frame-by-frame
- Display the annotated video in a GUI window Press q to quit the video playback.
Qt plugin error: Could not connect to display Make sure:
- Your X11 server is running
- DISPLAY is correctly set
- You’ve run xhost +local:docker Plugin not loading: xcb not found Ensure the following packages are installed in the container: libxcb1 libx11-xcb1 libxcb-glx0 libxcb-xinerama0 libxcb-render0 libxcb-shape0 libxcb-shm0 libxrender1 libxext6 libxkbcommon-x11-0 libxcb-cursor0
This project is open-source.
- MediaPipe
- OpenCV