Publications

R. Geirhos, Patricia Rubisch, J. Rauber, C. R. M. Temme, C. Michaelis, W. Brendel, F. A. Wichmann (0001). Inducing a human-like shape bias leads to emergent human-level distortion robustness in CNNs. Journal of Vision (19).

R. Geirhos, Patricia Rubisch, C. Michaelis, M. Bethge, F. A. Wichmann, W. Brendel (0001). ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness. In ICLR.

S. Otte, Patricia Rubisch, M. V. Butz (0001). Gradient-based learning of compositional dynamics with modular RNNs. In ICANN.

S. Haghiri, Patricia Rubisch, R. Geirhos, F. Wichmann, U. von Luxburg (0001). Comparison-based frameword for psychophysics: Lab versus crowdsourcing. arXiv.