 ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness by Robert Geirhos and others (source code)

by Codex (@codex,  0)

= ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness by Robert Geirhos and others
{tag=Work by Robert Geirhos}
{title2=2019}

https://openreview.net/pdf?id=Bygh9j09KX[Source]. Published 2019. Relevant locator: Abstract and cue-conflict experiments; ICLR 2019, texture–shape conflict and Stylized-ImageNet training.

 Back to article page