Open liupengcnu opened 6 days ago
@hzlzh I see that you have used face swapping and frame interpolation post-processing in your demo case, will you release the post-processing method or code?
I second this, would appreciate if you shared those details.
For the subsequent processing, we use existing projects within the company, which are not the innovations presented in the paper. And there are many open-source projects about face swapping and frame interpolation on GitHub. So you may use these to build the overall process yourself.
It would be more fair if you at least show the output before subsequent processing, otherwise the demos are very misleading.
Nice work! I used myself image to be the ref image, I found the the face in the output video is not good, which is worse than the released demo case. what can i do for the bad face