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

It does not work with custom poses or different poses other than the pair given in the example. #21

Open
adityapandey9 opened this issue Jul 3, 2024 · 3 comments

Comments

@adityapandey9
Copy link

It does not work with custom poses or poses other than the pair in the example.

I have used the image img3.png given in the images folder with pose2, I am getting the below output.

Screenshot 2024-07-03 at 1 09 10 PM

When I try to use a custom pose like the one below with any subject images, it does not work.

Screenshot 2024-07-03 at 1 11 00 PM

Any help would be appreciated.

@muzishen
Copy link
Collaborator

muzishen commented Jul 4, 2024

Thanks for your attention, the first case that fails may require multiple sample generation several times. Because we have limited data and computing power, but the stabilizing effect outside the domain requires more data and training steps. The problem in the second case seems to be that the posture is different from the one we used. It is suggested to use the posture detection tool we provided.

@adityapandey9
Copy link
Author

Hi @muzishen thank you for your response. So, do I need to retrain the model for each new subject? This is similar to the dreambooth, right?

@muzishen
Copy link
Collaborator

muzishen commented Jul 9, 2024

Hi @muzishen thank you for your response. So, do I need to retrain the model for each new subject? This is similar to the dreambooth, right?

Don't need it. As long as the number of ids is sufficient, then our ckpt can continue training.

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

2 participants