Is it possible to get the video stream at the browser level so that
I can do face recognition and OCR on the video feed?
You can take a look to some posts:
- Real-time image processing on the web | semi/signal
- Taking still photos with WebRTC - Web APIs | MDN
- video + canvas = magic | HTML5 Doctor
Also, latest Chrome version has a more efficient way to manipulate video data: