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

Generate Depth Images of TUM-RGBD Dataset #50

Open
czeyveli1 opened this issue Jan 30, 2025 · 3 comments
Open

Generate Depth Images of TUM-RGBD Dataset #50

czeyveli1 opened this issue Jan 30, 2025 · 3 comments

Comments

@czeyveli1
Copy link

Hello everyone,
First of all, congratulations on the amazing work!

I have a question about obtaining depth datasets for different datasets. I would like to generate depth images from the TUM-RGBD dataset. For instance, in the TUM-RGBD dataset, there is a scene called "fr1/desk," which contains 613 RGB images with a resolution of 640×480. How can I generate the corresponding depth dataset using your project? I couldn’t fully understand the process.

@haofeixu
Copy link
Member

Hi, thanks for your kinds words.

For depth prediction, you might want to check out the script in our unimatch repo: https://github.com/autonomousvision/unimatch/blob/master/scripts/depthsplat_depth_demo.sh

You can refer to the data loading example on scannet and modify it accordingly to load your data.

@czeyveli1
Copy link
Author

czeyveli1 commented Jan 31, 2025

Can I use the DepthSplat repo to generate a stereo depth model? Actually, I don’t just want to obtain a depth dataset—I can already do that using the DepthAnything model. Instead, I want to generate a scale-consistent depth dataset.

@haofeixu
Copy link
Member

haofeixu commented Feb 7, 2025

Hi, if you only want to generate depths with consistent scales, the unimatch demo code would be easier to use, and the dataloader is easy to set up.

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