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

Query where to find more data to train MiDiffusion? #8

Open
AnarchistKnight opened this issue Dec 5, 2024 · 1 comment
Open

Query where to find more data to train MiDiffusion? #8

AnarchistKnight opened this issue Dec 5, 2024 · 1 comment

Comments

@AnarchistKnight
Copy link

AnarchistKnight commented Dec 5, 2024

Thanks again to the authors for sharing their code.

Do you have any idea, besides 3DFront, where to find more indoor room layout to train the model? A simple motivation behind asking this question is that, more and better data surely means better quality of the model output.

As I have some experience in 3d reconstruction and neural rendering, I know reconstructing indoor scenes from rgb or rgbd sequences is problematic. For example, rgbd scenes might be synthesis and of low quality. And the number of frames in a sequence might not be enough to reconstruct a compete indoor scene.

I would be very happy if you are willing to discuss this.

@SiyiHu
Copy link
Collaborator

SiyiHu commented Dec 18, 2024

You would need a dataset with at least object localization, object labels, and floor plans.

I think Matterport3D provides all required data, but it only contains 90 building scenes, including a few non-residential scenes. I have also used HM3DSem for other projects. This is a larger and more recent dataset. It should have floor plans but I am pretty sure it does not contain room labels. You can probably annotate rooms yourself and use objects as heuristic to sort different types of rooms.

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