Open tyrink opened 8 months ago
The paper claims to use 3DMM to get the facial keypoints, however, the code implementation directly uses the 2D landmarks of the target face detected by the dlib detector, which is obviously different.
The paper claims to use 3DMM to get the facial keypoints, however, the code implementation directly uses the 2D landmarks of the target face detected by the dlib detector, which is obviously different.