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 to add mask supervision for object 3d reconstruction ? #2

Open
Jiakui opened this issue Sep 18, 2024 · 1 comment
Open

How to add mask supervision for object 3d reconstruction ? #2

Jiakui opened this issue Sep 18, 2024 · 1 comment

Comments

@Jiakui
Copy link

Jiakui commented Sep 18, 2024

Dear author,

How to add mask supervision for object 3d reconstruction ?

Thanks

@yyfz
Copy link
Collaborator

yyfz commented Sep 18, 2024

Hi, thank you for your interest in our work.

Please refer to this link for guidance on how to read mask images in your dataparser. You can use this as a reference to incorporate the code snippet for reading masks into your own dataparser and set include_foreground_mask to True.

Additionally, in our model, you can find the original mask supervision code belonging to sdfstudio/nerfstudio at this location. Simply set fg_mask_loss_mult to a value greater than 0.

It's worth noting that we have not yet experimented with using masks for supervision. Therefore, if possible, it would be advisable to remove the mask portion from the unbias loss at this location, as this part of the unbias loss is meaningless.

In summary, once your dataparser is ready to read mask images and your model is prepared, you only need to make the following adjustments to your original startup script:

ns-train ... --pipeline.model.fg_mask_loss_mult <Coefficient of mask supervision> ... <your-data-parser (for example `sdfstudio-data`> ... --include_foreground_mask True

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