Welcome to the Face Reconstructor project! π This advanced application utilizes the cutting-edge capabilities of the Mediapipe library Holistic tool
to accurately identify and analyze a comprehensive set of 468 facial landmarks. Leveraging the power of OpenCV (cv2), this system employs sophisticated
algorithms to generate a meticulously crafted triangulation mesh, wherein each polygon is expertly filled with a seamless patch of skin derived from
the original frame. πΈπ‘
πβ¨ Facial Landmark Detection: Utilizing Mediapipe Holistic, this system accurately detects key facial landmarks, revealing detailed structural insights. π―π‘
π³β¨ Triangulated Mesh Generation: With OpenCV (cv2), this application creates a precise triangulation mesh, enabling advanced analysis and visualizations of facial geometry. ππ²
π¨β¨ Patch-based Skin Reconstruction: This system achieves lifelike face reconstruction by filling triangulation mesh polygons with carefully selected skin
patches, maintaining a natural appearance. ποΈπ§βπ¨
- π΄ Fork the repository to your GitHub account.
- β¬οΈ Clone the forked repository to your local machine.
- πΏ Navigate to the local repository using the command line.
- π» Run the script using the following command:
python reconstructor.py --input_image assets/example_video.mp4 --output_path results/processed_example_video.mp4,
where assets/example_video.mp4is the path to the input video you want to extract the face from. - βοΈ The generated mesh will be saved in the "results" folder for further usage.
The following dependencies are required to run the Face Reconstructor:
- Python 3.x
- Mediapipe
- OpenCV Python
- Numpy
You can install these dependencies manually using the package manager of your choice.