Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Not doing proper face swap if the he target face have moving hand #669

Open
jaikargupta opened this issue May 7, 2024 · 5 comments
Open

Comments

@jaikargupta
Copy link

Not doing proper face swap if the he target face have moving hand
Please find the attached video.

roop3hjgfhjgjh.1.mp4
@lysxelapsed
Copy link

This should work with clip2seg:

  • Open up Advanced Masking (below the input face)
  • Select Clip2Seg as Face masking engine
  • Enter hand in the field below

But in my experience, the result won't be perfect even with masking.

@jaikargupta
Copy link
Author

This should work with clip2seg:

  • Open up Advanced Masking (below the input face)
  • Select Clip2Seg as Face masking engine
  • Enter hand in the field below

But in my experience, the result won't be perfect even with masking.

Yes, I tried but the result is same :(

@C0untFloyd
Copy link
Owner

C0untFloyd commented May 10, 2024

Do you still have the original untouched clip? Perhaps upload this here so I can give it a go. I think xseg should work well with this but the real problem is if the hand covers a lot of the face there will be NO face detected and therefore no face swapping/masking will happen.

@jaikargupta
Copy link
Author

Do you still have the original untouched clip? Perhaps upload this here so I can give it a go. I think xseg should work well with this but the real problem is if the hand covers a lot of the face there will be NO face detected and therefore no face swapping/masking will happen.

Yes, I still have and here it is

3nghj.mp4

@C0untFloyd
Copy link
Owner

Non cherry picked quick run using DFL XSeg und GPEN:
https://github.com/C0untFloyd/roop-unleashed/assets/131583554/9d97399f-31c9-4c69-bad8-fd33582dbe04

I think this is better than your try, it doesn't have any half transparent eyes on the hands or stuff like that, so the masking is working very well. However this is still far from perfect as the face detection has problems to get the face orientation with just 1 eye. This is as good as it gets without training I'm afraid.
My Settings:
settings

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants