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

Some wired pink-green artifact #8

Open
YotamWolf opened this issue Feb 14, 2023 · 11 comments
Open

Some wired pink-green artifact #8

YotamWolf opened this issue Feb 14, 2023 · 11 comments

Comments

@YotamWolf
Copy link

Hi thanks for the amazing project, the best I've seen so far.

On some of the 3D models I'm trying I receive som weird pink-green artifact like salt and pepper noise, any idea why is that
final_obj_7bc0a1e4-ac51-11ed-a5e4-06e882d545ae
happening?
Screenshot 2023-02-14 at 13 19 47

@AmanKishore
Copy link

I'm getting a similar issue:
image

@pengwangucla
Copy link

+1

@YuliangXiu
Copy link

+1

image

@qzq2514
Copy link

qzq2514 commented Apr 6, 2023

Differentiable rendering has reconstruction error under the api of kaolin,this is the main reason to cause this green artifact. Pink artifact is caused by its inaccurate method to calculate trimap.

@pengwangucla
Copy link

Differentiable rendering has reconstruction error under the api of kaolin,this is the main reason to cause this green artifact. Pink artifact is caused by its inaccurate method to calculate trimap.

any way to fix these issue ? e.g. for the first issue, should we use pytorch3d differential renderring or there is a corrected way.

@qzq2514
Copy link

qzq2514 commented Apr 23, 2023

Differentiable rendering has reconstruction error under the api of kaolin,this is the main reason to cause this green artifact. Pink artifact is caused by its inaccurate method to calculate trimap.

any way to fix these issue ? e.g. for the first issue, should we use pytorch3d differential renderring or there is a corrected way.

I haven't try pytorch3d differential renderring, but I think this error may still exist for any texture reconstruction under differentiable rendering, but you can try to map the view image to texture space directly, just like the inverse of kaolin's texture_mapping

@reynoldscem
Copy link

I am also experiencing this. Has anyone found a solution?

@YuzhiChen001
Copy link

Differentiable rendering has reconstruction error under the api of kaolin,this is the main reason to cause this green artifact. Pink artifact is caused by its inaccurate method to calculate trimap.

"inaccurate method to calculate trimap",I want to know why the author's method is inaccurate?How do you think it should be calculated?thanks.

@qzq2514
Copy link

qzq2514 commented Sep 1, 2023

Differentiable rendering has reconstruction error under the api of kaolin,this is the main reason to cause this green artifact. Pink artifact is caused by its inaccurate method to calculate trimap.

"inaccurate method to calculate trimap",I want to know why the author's method is inaccurate?How do you think it should be calculated?thanks.

author's method to get uncolored mask is to calculate difference between textured uvmap and base color, this will be mostly affected by the color of texture. The correct method is to use reverse process of rasterizing.

@smha1012
Copy link

smha1012 commented Sep 8, 2023

This problem occurs because the mesh color cannot be optimized perfectly in trainer.py, i guess. However, I think this is not a fundamental reason and it seems closely related to the normal parameters from the mesh(=self.mesh_model.get_params() in Adam optimizer) and the train/eval_grid_size configs.

Therefore, since each person uses different kinds of mesh, it will be very difficult to find a perfect solution.

@qzq2514
Copy link

qzq2514 commented May 23, 2024

Differentiable rendering has reconstruction error under the api of kaolin,this is the main reason to cause this green artifact. Pink artifact is caused by its inaccurate method to calculate trimap.

"inaccurate method to calculate trimap",I want to know why the author's method is inaccurate?How do you think it should be calculated?thanks.

author's method to get uncolored mask is to calculate difference between textured uvmap and base color, this will be mostly affected by the color of texture. The correct method is to use reverse process of rasterizing.

We have in part solved this problem, you can find this in https://github.com/OpenTexture/Paint3D

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

8 participants