This project implements a pose tracking system using MediaPipe for real-time pose estimation. The system supports both webcam and RealSense cameras, allowing for both 2D and 3D visualization of tracked landmarks.
The pose_estimation.py
script serves as the main entry point for the application, initializing the camera, processing video frames, and visualizing pose landmarks. The pose_visualiser.py
script provides a 3D visualization environment for the detected pose points and camera orientation.
- Real-time pose tracking using MediaPipe
- Support for webcam and RealSense cameras
- 2D and 3D visualization of pose landmarks
- Depth awareness when using depth-capable cameras
- Interactive visualization with mouse controls
To run this project, ensure you have the following dependencies installed (requirements.txt). In some versions there is a backend conflict between Pyqt5
and opencv
, therefore i have mentioned specific versions for the same.
To start the pose tracking application, run the following command in your terminal:
python pose_estimation.py --cam-type <realsense|webcam> --depth --model <lite|full|heavy> --width <width> --height <height>
Arguments
--cam-type:
Specify the type of camera to use (default isrealsense
).--depth:
Enable depth processing (only applicable for depth cameras).--model:
Choose the model type for pose estimation (default islite
).--width:
Set the width of the video frame (default is640
).--height:
Set the height of the video frame (default is480
).
Example
python pose_estimation.py --cam-type webcam --model lite --width 640 --height 480
This project is licensed under the MIT License - see the LICENSE file for details.
- This project uses MediaPipe for pose estimation.
- 3D visualization is powered by Vispy.