ocastan

Results 1 comments of ocastan

IMHO this is due to MTCNN weights trained with RGB images whereas OpenCV use BGR: changing image with image[..., ::-1] in L62 in infer_on_video.py or L55 in face_verify.py, give me...