The place to discuss Faceswap and Deepfakes
https://forum.faceswap.dev/
Hey,
I had a question about training profile faces and their alignments.
Should I be manually deleting the whole eye landmarks for his right eye? Or does the AI understand what's going on?
Appreciate feedback.
You can't delete the "right eye". All 68 Landmarks need to be present. Therefore for profile shots the hidden landmarks will tend to "stack" onto the side of the face.
As long as the face image looks aligned, this is fine.
Thanks torzdf. Could you take another look at this preview snapshot and perhaps give me your best guess as to whats going on?
I'm fairly happy with 4/5 previews, its just that one in the middle that is not turning out right. I haven't disabled "No Warp" yet, as I'm waiting for the previews to turn out all satisfactory before disabling.
The source is a 1080p video (torrent, not youtube), training on an RTX2080, realface trainer. 225k iterations.
The things that stand out to me are that;
1) it's a profile face, most previews of the profile face seem to be turning out this way.
2) the lighting is a bit dark, but not sure it's TOO dark?
Any possible solutions are appreciaited.
Profiles are always tough and often require postprocessing to get the best result. The problem is that you're blurring the edges and the edges in that case has a new nose sticking out past the old one. That's getting blurred with the background. There is no automatic solution for something like that. Instead you'd want to composite it manually so you can handle that (literal) edge case.
thanks for your reply bryan, when you say composite it manually, are you talking about fixing the alignments of the nose? Or doing something to make the source video more clear?
If it's the former, does something like this do the job or am I a complete moron?
original alignment: https://ibb.co/mJnwtNK
simply changed landmark "31" to the edge of his nose.
correction alignment: https://ibb.co/NrykgBs
appreciate your help.
No compositing is the process of combining two images into one. Compositing in my example would mean using video editting software like After Effects to combine the images. You would probably need to do it frame-by-frame in order to get the best results.