About the category list of ImageNet-100 subset

[szq0214]

Hi @HobbitLong, thanks for your great work and also sharing the code. I guess the ImageNet-100 is not a conventional subset so I wonder if you can share the list since we also don't have enough resources to run on the full ImageNet ==.

[ghost]

Same issue：I have tried the tiny-imagenet-100-A dataset from https://tiny-imagenet.herokuapp.com/, is it same as the ImageNet-100 metioned in the README?

Btw, I follow the guide to train the dataset using MoCo, but the test_acc top1 is about 2.5%, and the train_acc top 1% is about 59%, I don't know which part of the exp maybe be wrong. Could you show some val dataset format or test_set acc curvey?

Thanks you，appreciate this job ~

[HobbitLong]

Hi, please find the list below:

n02869837 n01749939 n02488291 n02107142 n13037406 n02091831 n04517823 n04589890 n03062245 n01773797 n01735189 n07831146 n07753275 n03085013 n04485082 n02105505 n01983481 n02788148 n03530642 n04435653 n02086910 n02859443 n13040303 n03594734 n02085620 n02099849 n01558993 n04493381 n02109047 n04111531 n02877765 n04429376 n02009229 n01978455 n02106550 n01820546 n01692333 n07714571 n02974003 n02114855 n03785016 n03764736 n03775546 n02087046 n07836838 n04099969 n04592741 n03891251 n02701002 n03379051 n02259212 n07715103 n03947888 n04026417 n02326432 n03637318 n01980166 n02113799 n02086240 n03903868 n02483362 n04127249 n02089973 n03017168 n02093428 n02804414 n02396427 n04418357 n02172182 n01729322 n02113978 n03787032 n02089867 n02119022 n03777754 n04238763 n02231487 n03032252 n02138441 n02104029 n03837869 n03494278 n04136333 n03794056 n03492542 n02018207 n04067472 n03930630 n03584829 n02123045 n04229816 n02100583 n03642806 n04336792 n03259280 n02116738 n02108089 n03424325 n01855672 n02090622

Testing curve of MoCo:

[HobbitLong]

@szq0214 @YanruoYang . For MoCo on ImageNet-100, setting --alpha 0.99 gives slightly better results (shown above) than --alpha 0.999, which gives an accuracy of around 72.0%.

[ChongjianGE]

@HobbitLong
Hi, thank you for providing the detailed list.
Actually, there are 1300 images in one single sub-file. I wonder whether you use all of the images in ImageNet-100 subset or not?

[qsunyuan]

I found two versions of Imagenet100 dataset.
I was so confused which one should I choose.

Vesions 1:
https://www.kaggle.com/ambityga/imagenet100 or https://github.com/arthurdouillard/incremental_learning.pytorch/tree/master/imagenet_split

Vesions 2:
as disscussed above. the label index.

Hope to get ur suggestions early.

Thx.

[HobbitLong]

Hi, @qsunyuan , I think either one is OK. If you want to directly compare with numbers in our CMC paper, you could use the one this repo provides.