AliaksandrSiarohin

For example, yes avatarify.

I never tried this, I guess tricks with replacing standard blocks with mobile net blocks should not hurt the performance much.

I presume that 68 keypoints means supervised keypoints obtained from face-aligment lib or dlib? You need to modify the code in this case.

Did you adjust a batch size? You should set batch size to 5 or even less.

Yes, dataset is needed

Taichi have very poor Generalization, so basically it only works with other Taichi videos.

Try it out.

You can try to add find best frame in demo.

You either did not run first or second cell, run them again.

@HashedViking Never did this, but since there is no runtime branching I guess it should be possible.