This project is a test implementation of extracting unique faces detected on YouTube videos with a given URL. I have not implemented CLI tools or video ID extraction. It is a plain coded URL that you must change or extend functionality to deliver new video URLs or process batch videos.
The core of this project is built on top of the face_recognition and rembg Python libraries, which are doing pretty much all of the work; I just made a composable reading from a given video file using Python OpenCV.
This project doesn’t have external service dependencies, and everything is inside an [experiment.py](http://experiment.py) file, where it will handle the video download using the pytube library and write extracted frames inside the same directory.
pip install -r requirements.txt
python experiment.pyThis project doesn’t have that much in it, because it was very small functionality that I am building for one of my products within [treescale.com](http://treescale.com) as a Face detection API Endpoint.