Title | ||
---|---|---|
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness. |
Year | Venue | DocType |
---|---|---|
2019 | International Conference on Learning Representations | Conference |
Citations | PageRank | References |
0 | 0.34 | 0 |
Authors | ||
6 |
Name | Order | Citations | PageRank |
---|---|---|---|
Robert Geirhos | 1 | 67 | 5.21 |
Patricia Rubisch | 2 | 33 | 1.09 |
Claudio Michaelis | 3 | 0 | 1.01 |
Matthias Bethge | 4 | 2 | 4.12 |
Felix A. Wichmann | 5 | 0 | 0.68 |
Wieland Brendel | 6 | 0 | 2.70 |