Notebooks featuring use of the face_alignment library for processing videos and for obtaining input landmarks for the project Deep3DFaceReconstruction. Best to view and run these Notebooks in Google Colab as in that case there should not occur any dependency conflicts and installation of all required packages should go smoothly.
Performs face alignment (i.e., detection of 68 facial landmarks) on each frame of a given video. For processing a video, ffmpeg is used. Related Colab Notebook.
Performs face alignment and extracts 5 landmarks needed to run the project Deep3DFaceReconstruction. Related Colab Notebook.
Thanks to Adrian Bulat for the development of the amazing face_alignment library. Also, thanks to the authors of the Deep3DFaceReconstruction project for making their work publicly available.