search

SHI-Labs / OneFormer

OneFormer: One Transformer to Rule Universal Image Segmentation, arxiv 2022 / CVPR 2023
https://praeclarumjj3.github.io/oneformer
MIT License
1.44k stars 130 forks source link

inquery about ade20k caption label #8

Closed Mollylulu closed 1 year ago

Mollylulu commented 1 year ago

Hi, thanks for your great work. I did not find the caption-related annotations on the ADE20K website, could you point out how to get them? or how to generate them?

praeclarumjj3 commented 1 year ago

Hi @Mollylulu, thanks for your interest in our work.

We do not use captions to train or test OneFormer. Their mention was there from a previous experimental setting. I have removed them from the file list. Thanks for pointing it out to me.

Mollylulu commented 1 year ago

noted, thanks :)