Question about data augmentation.
Talegqz opened this issue · comments
Thanks for your excellent work.
When I read the code, I found that you use data augmentation (color jitter), which is not mentioned in the paper.
So I want to know how much this module plays.
Could it be generalized without this module?
Thanks for your reply!
Hello!
I used color jittering because the number of subjects in the dataset is small.
From my previous experiences, using the color jittering can help better generalize to unseen subjects given the small number of subjects in the dataset.
But I don't have results without the jittering module for this dataset, so you can try it by yourself :) You can do it by simply changing the 'jitter' argument to False in the provided training command.
Hope this can help you!