Open roberthobblebottom opened 3 months ago
Briefly looking at XGBClassifier() for AU detection, increasing the number of jobs through the parameter n_jobs
seems like an option.Or even -1 n_jobs
value to use all CPU workers? Has this been tried before? we may try this out if we aregoing to pause on our DeepFace emotion inference running right now.
It took too long to run my 4k 30fps video frame by frame.
we also tried just doing detect_facebox(), detect_landmarks() and detect_aus(). It is still taking too long.
we also tried openface2 but I have so much problems with the docker image, it has been frustrating.
we fell back to DeepFace to just do emotion recognition and it is taking along time too.
But we thank all open source contributors, without them there will not be libraries in the first place to test out.