OpenGVLab / InternImage

[CVPR 2023 Highlight] InternImage: Exploring Large-Scale Vision Foundation Models with Deformable Convolutions
https://arxiv.org/abs/2211.05778
MIT License
2.47k stars 231 forks source link

Typo in classification/meta_data/map22kto1k.txt? #278

Closed ZhiqiGao2001 closed 6 months ago

ZhiqiGao2001 commented 7 months ago

Hello! Thanks for the open-source model, I am doing some research on it and find a possible typo

image I think there should not be a duplicate in the mapping. The wrong one (number 850) should be [n04399382: teddy bear] from what I see in the other metadata file in the Imagenet-1k repo? (but n04399382 does not exist in the 22k_class_to_idx.json?)