Use aligned features from MTCNN instead of dlib

kaschmo / sh_face_rec

Face recognition system for any IP Camera and openHAB

30 stars 10 forks source link

Use aligned features from MTCNN instead of dlib #6

Open kaschmo opened 6 years ago

kaschmo commented 6 years ago

Currently running MTCNN for face detection and then dlib for crop and align. MTCNN returns alignment features that need to be converted but could be passed to CNN for face embedding generation