"Named as TIN-597, the resulting dataset has 597 classes which do not overlap with any of the categories from test OOD and serves as a good candidate for train OOD."
When I follow above description, I found 682 classes.
You can find the class names (wordnet IDs in ImageNet) if you download the benchmark_imglist.ziphere and go to benchmark_imglist/cifar10/train_tin597.txt. You can then see the class names such as n03710637.
Thank you for your benchmark dataset.
I just want to know the all of 597 classes name.
"Named as TIN-597, the resulting dataset has 597 classes which do not overlap with any of the categories from test OOD and serves as a good candidate for train OOD."
When I follow above description, I found 682 classes.
How can I get them ?