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

render_pkg["median_depth"] with artifacts vs render_pkg["expected_depth"] #51

Open
maorp opened this issue Sep 2, 2024 · 5 comments
Open

Comments

@maorp
Copy link

maorp commented Sep 2, 2024

Hi, first thanks for the nice work! I'm trying to use the depthmap directly and i see this difference between "median_depth" and "expected_depth" - which is much smoother and with less jumps but has "tails" of points to the background. see images below:
(these are pointclouds created from the depthmaps)
image
image

@maorp maorp closed this as completed Sep 8, 2024
@renhaofan
Copy link

@maorp Could you let me know the reason?

@maorp
Copy link
Author

maorp commented Sep 10, 2024

I still dont know for sure. It seems from the code that median depth is computed per pixel using ray intersection and not alpha blending. This creates a better multi view consistent depthmap but with those weird band jumps. Its not clear if its by design or its a bug.

@maorp maorp reopened this Sep 10, 2024
@BaowenZ
Copy link
Owner

BaowenZ commented Sep 10, 2024

Hi, the depth remains continuous within a kernel but experiences jumps when the median kernel changes.

@maorp
Copy link
Author

maorp commented Sep 10, 2024

Thanks! Is there a way to control it or improve it? Im happy to help coding it if needed

@BaowenZ
Copy link
Owner

BaowenZ commented Sep 11, 2024

I can't think of a complete solution. However, the "stop the pop" proposed a better method to determine the order of kernels. Maybe it can improve the accuracy of median depth.

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

3 participants