imagenet-a

[jieweilai]

hello,i have a quession,if you test your model in imagenet-a directly?i find the classes in imagenet-a is different from imagenet-200?how do you slove it?

[LiYingwei]

For the models trained on ImageNet-200, we don't test them on ImageNet-A. We only test the ImageNet-1000 trained model on Imagenet-A. This script tells you how to do it.

[jieweilai]

thanks for your reply,i thanks you very much!

…

------------------&nbsp;原始邮件&nbsp;------------------ 发件人: "LiYingwei/ShapeTextureDebiasedTraining" <notifications@github.com&gt;; 发送时间:&nbsp;2020年11月24日(星期二) 晚上9:09 收件人:&nbsp;"LiYingwei/ShapeTextureDebiasedTraining"<ShapeTextureDebiasedTraining@noreply.github.com&gt;; 抄送:&nbsp;"『☆初陽卍綻放』"<2556004343@qq.com&gt;;"Author"<author@noreply.github.com&gt;; 主题:&nbsp;Re: [LiYingwei/ShapeTextureDebiasedTraining] imagenet-a (#3) For the models trained on ImageNet-200, we don't test them on ImageNet-A. We only test the ImageNet-1000 trained model on Imagenet-A. This script tells you how to do it. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.