Closed luiscosio closed 5 years ago
video doesnt satisfy me. Result trained on existing 68 points predictor, we can see 3d face mask jittering. So why to create predictor based on predictor? This is data loss.
Bump, because the developement seems quite active and they now stated texture editing (which could help us to get better eye details) and face swapping as possible options
You are right @tjess78 .
Have you seen this @iperov ?
looks crap. Also video can produce awful result.
Have you guys seen this?
Repository:
https://github.com/YadiraF/PRNet
Relevant discussion:
https://news.ycombinator.com/item?id=16977516 https://www.reddit.com/r/MachineLearning/comments/8gk1rt/p_joint_3d_face_reconstruction/