Anybody reproduced the similar results?

[junyongyou]

I have downloaded the dataset, including the real image sets used in the paper. I tried to train ConvneXt-S following almost the same settings as the paper. However, I only get around 50% accuracy on both the train set and the val set, meaning that it equals to a random guess. Has anybody tried to use the dataset and trained a classifier with decent accuracy? Thanks.

[iamwangyabin]

no