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

How does the pre-trained weights work on a single person, like Seth? #8

Open
shengyuting opened this issue May 9, 2024 · 5 comments

Comments

@shengyuting
Copy link

Thank you for your great work. I have trained the model by the open dataset which described in the paper. However, in my test results, the image texture is not clear. Whether it can be greatly improved by fine-tuning training or there is no such problem in your test results? Looking forward to your reply very much!!!

@bone-11
Copy link
Collaborator

bone-11 commented May 10, 2024

Thank you for your attention!
The pre-trained weight indeed generates unclear image texture, which can be attributed to training on various appearances leading the model to learn mean features. In our work, fine-tuning on specific appearances can effectively enhance the quality of the appearance. Furthermore, hyperparameter settings, such as CFG and controlnet weight, also impact the results.

@andrewkuo
Copy link

Thanks for your work! Relating to this, do you have any timeline for when you will release the training and preprocessing code?

@bone-11
Copy link
Collaborator

bone-11 commented May 10, 2024

Thanks for your interest. I'm trying my best to release the relevant code and pretrained weights within this month.

@nerlfield
Copy link

Hi, thank you for this great work! I wanted to ask if there are any updates on the training and fine-tuning code release?

@kike-0304
Copy link

Thanks for your interest. I'm trying my best to release the relevant code and pretrained weights within this month.

When will you open source?

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

5 participants