Open liliangqi opened 6 years ago
The detection and alignment models are copied from MTCNN. The recognition model is trained on MS-Celeb-1M. Our training code would not be released for some copyright reasons. But I think it is a very simple training process by just classifying the faces by their identities.
Hi, I'd like to know which dataset did you use to train such a model. And is there any codes for the training step?