Provide detailed data augmentaion setup or codes mentioned in paper?
LightningChan opened this issue · comments
LightningChan commented
Hi, I think you add lots of data augmentation to the validation set. This results in a large gap between the training and validation sets. This can lead to a situation where a model performs well, possibly simply due to a closer augmentation setup used in the validation set. It is very unfair when compared methods on the validation set. So, I suggest you provide detailed data augmentation setup or codes mentioned in the paper.
Yinan He commented
We used albumentations.
Davide Coccomini commented
Can you provide the exact augmentations used in the three setups?