Open LuciusPertis opened 1 year ago
there can be a more optimized way of doing this combining face-detect pipeline with pose-embedding
MAIN pipeline INP : preprocessed video file; no face database; maybe some incentives on cropping OUT : face_frame.obj; should contain details of a single continuous frame details of a face on-scene; imp. method to generate raw frames out of it; also collab with other face_frame.obj and efficient searching/pose-indexing