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

ADE20K-outdoor #23

Open
EndlessSora opened this issue Apr 18, 2022 · 1 comment
Open

ADE20K-outdoor #23

EndlessSora opened this issue Apr 18, 2022 · 1 comment

Comments

@EndlessSora
Copy link

Hi @SushkoVadim @edgarschnfld ,

Thanks for the excellent work. Starting from SPADE, I noted that a series of papers provided the benchmark results on the ADE20K-outdoor dataset, a subset of ADE20K that only contains outdoor scenes, used in Qi et al. In their repo, I only found the 1,035 ADE20K-outdoor test labels their provided 'result.zip' file.

Could you please share how I can find the list/data of ADE20K-outdoor and how to train the model (train: full, test: subset; Or: train: subset, test: subset) to make a fair comparison? I ask this question only because I wish to test my model on ADE20K-outdoor.

Thank you very much for any help you may provide.

@SushkoVadim
Copy link
Contributor

Hi,

I am not sure there exists an official list of the train/val splits for the ADE20K-outdoor dataset.

To construct this dataset, we selected images in the respective splits of full ADE20K, by filtering images containing indicative classes like "sky", "mountain", "earth", and filtering out obvious indoor classes like "ceiling", "bed", "sink". After this step, we manually skimmed through the selected images and manually filtered out few remaining indoor scenes. This resulted in the train/val splits closely matching the number of images reported in other papers. (very close to 10k and 1k, as far as I remember).
You can find the full list of ADE20K classes and their IDs here.

Good luck!

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