Ways to improve on the facial landmark detection model
akshay-hazari opened this issue · 5 comments
Is clmtools used to fetch facial landmarks here. It uses dlib but dlib results in a lot better accuracy in detecting the landmarks. There are a few codes in python and c++ with dlib. clmtools uses dlib.
This was something I was trying to figure to convert it to javascript https://github.com/kripken/emscripten/ . Is there way to convert or get something like clmtools, dlib, to detect facial landmarks and convert the detecting part to javascript.
@akshay-hazari Did you find a solution for that please?
I would definitely be interested in improving clmtrackr. Does anyone know how I can use other libraries and tie in some of the key points for facial tracking to assist clmtrackr or takeover?
@lakano @msj121 I had been trying. Here are the issues I was facing then. emscripten-core/emscripten#5435 . First I had emscripten build opencv to js, then dlib to js . Then have them linked. But it got stuck at the same point. I guess you can give it a try and let me know as well.
Does anyone has converted dlib to js? I am looking for dlib library in js for facial landmark detection in real time video stream.
Same problem here. clmtracker doesn't detect eye opening and closing.