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

Questions about process custom datasets #6

Closed
ZhengyuDiao opened this issue Sep 13, 2024 · 5 comments
Closed

Questions about process custom datasets #6

ZhengyuDiao opened this issue Sep 13, 2024 · 5 comments

Comments

@ZhengyuDiao
Copy link

ZhengyuDiao commented Sep 13, 2024

Thanks for release the great work!
while I use this repo to process my custom dataset which contains multi-view videos and camera calibrations. I rewrite the dataset codes and try to keep dataloader ouput in same format with nersemble_dataset.py, however, the alignment result below looks strange. the photo loss is inf in process.
Are there any other things to pay attention to when processing custom dataset?
截图 2024-09-13 21-14-46

@ShenhanQian
Copy link
Owner

Hi, thanks for sharing your result! How many views do you have in your multi-view setting?

@ZhengyuDiao
Copy link
Author

ZhengyuDiao commented Sep 14, 2024

Hi, thanks for sharing your result! How many views do you have in your multi-view setting?

I use 5 views in my experiment.

@ShenhanQian
Copy link
Owner

I see. Since you have fewer views than NeRSemble (16 views), you may need to weaken the regularization on the texture map by --w.reg_tex_tv 1e4 or 3e4. The texture map after optimization should be a little blurry than the GT image but much sharper than what you have now.

@ZhengyuDiao
Copy link
Author

Thanks!

@ShenhanQian
Copy link
Owner

Which value for w.reg_tex_tv works best for you? Do you need to do other adjustments to make it work?

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