Alibaba-MIIL / ImageNet21K

Official Pytorch Implementation of: "ImageNet-21K Pretraining for the Masses"(NeurIPS, 2021) paper

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Normalize

akuxcw opened this issue · comments

Thank you for this awesome work! I want to make sure that you only rescale the input image to 0-1 instead of normalizing them by (img-mean)/std like we usually do in ImageNet training. Is that right?

commented

correct.
the mean,std normalization doesn't contribute anything in practice

Thanks! Just a piece of advice, you may declare this normalization strategy somewhere. Because it's important to align the input scales for the downstream tasks.