118 citations · 319 across the 13 of their papers we have counts for
Showing 2018Show all
2 papers · 1 filter
cs.CV2018
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness
Robert Geirhos, Patricia Rubisch, Claudio Michaelis +3
Convolutional Neural Networks (CNNs) are commonly thought to recognise objects by learning increasingly complex representations of object shapes. Some recent studies suggest a more…
cs.CV2018
Generalisation in humans and deep neural networks
Robert Geirhos, Carlos R. Medina Temme, Jonas Rauber +3
We compare the robustness of humans and current convolutional deep neural networks (DNNs) on object recognition under twelve different types of image degradations. First, using thr…