FaceMesh dataset creating tool Using google's mediapipe

How to use it.

This reposiry contans uses mediapipe model to extract face mesh points from a video file or webcam. The extract points are both 2D(x, y points) and 3D(x, y, z points) points. The extracted mesh is saved with json file format for each frames of the video or webcam feed. If the points are extracted from webcam, the video will also be saved in data/webcam.avi.
If you want to just visiualize the face mesh without saving the facemesh points, you can also use the code in drawFaceMesh.py file. Here is how you can easily visualize the facemesh using this script.
Using video file

$ python drawFaceMesh.py --video_path path/to/video.mp4

Using the webcam

$ python drawFaceMesh.py --camera

To save the facemesh points extracted from each frames of the video or webcam, makeDataset.py file.

From video file, the extracted points will be saved inside data/detected_mesh/videoname folder.

$ python makedDataset.py --video_path path/to/video.mp4

From the webcam, the extracted points will be saved inside data/detected_mesh/webcam folder and the video will be saved in data/webcam.avi file

$ python makeDataset.py --camera

File struecture structure:

.
├── ./data
│   └── detected_mesh
│       ├── videoname
│       └── webcam
├── ./log
├── ./utils
│   ├── __init__.py
│   ├── arguement.py
│   └── utils.py
├── cameraFeed.py
├── drawFaceMesh.py
├── faceMesh.py
├── makeDataset.py
├── videoFeed.py
└── README.md

Dependency

mediapipe==0.8.9.1

$ pip install mediapipe

Resources

Mediapipe