Closed frankier closed 3 years ago
face_recognition/dlib has a pipeline like so: coarse face detection -> keypoint detection -> crop/scale to 150x150 -> embedder
We already have exactly the keypoints needed when we run with face keypoints https://github.com/CMU-Perceptual-Computing-Lab/openpose/issues/463
So the issue is just to put these into a format dlib can work with.
face_recognition/dlib has a pipeline like so: coarse face detection -> keypoint detection -> crop/scale to 150x150 -> embedder
We already have exactly the keypoints needed when we run with face keypoints https://github.com/CMU-Perceptual-Computing-Lab/openpose/issues/463
So the issue is just to put these into a format dlib can work with.