ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness - researchr publication authors

researchr

You are not signed in
Sign in
Sign up

Robert Geirhos, Patricia Rubisch, Claudio Michaelis, Matthias Bethge, Felix A. Wichmann, Wieland Brendel. ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness. In 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019. OpenReview.net, 2019. [doi]

This author has not been identified. Look up 'Robert Geirhos' in GoogleThis author has not been identified. Look up 'Patricia Rubisch' in GoogleThis author has not been identified. Look up 'Claudio Michaelis' in GoogleThis author has not been identified. Look up 'Matthias Bethge' in GoogleThis author has not been identified. Look up 'Felix A. Wichmann' in GoogleThis author has not been identified. Look up 'Wieland Brendel' in Google

runs on WebDSL