search

zhanglonghao1992 / One-Shot_Free-View_Neural_Talking_Head_Synthesis

Pytorch implementation of paper "One-Shot Free-View Neural Talking-Head Synthesis for Video Conferencing"
Other
764 stars 143 forks source link

About rotationt replacing the Jacobians #3

Closed zbdehh closed 2 years ago

zbdehh commented 3 years ago

Very good result. When I tried to reproduce this paper, I found that the Jacobian matrix was cancelled in the v3 version of arxiv. I canceled the Jacobians and got very worried results. I would like to ask if you use the Rotation parameter to replace the Jacobian. Also want to know if the code will be open source in the near future?

zhanglonghao1992 commented 3 years ago

@begins233 I'm trying training without the Jacobians. It looks fine so far.

zbdehh commented 3 years ago

@zhanglonghao1992 Thanks. There may be some bugs in my code, I will check the code again. This is my current result. The left is the output, the middle is the output with no expression coefficients, and the right is GT. Have you ever encountered such a situation where the posture is correct but the face is expressionless, I hope you can give me some suggestions, thank you.

https://user-images.githubusercontent.com/24779371/125877738-74d08d30-103d-4f16-88f4-bdd2b0d692c5.mp4

zhanglonghao1992 commented 3 years ago

@begins233 It looks like there's something wrong with your implementation.

zbdehh commented 3 years ago

@begins233 It looks like there's something wrong with your implementation.

Thank you, I try to keep checking my code. Hope you can open source as soon as possible.