A Self-Supervised Feature Map Augmentation (FMA) Loss and Combined Augmentations Finetuning to Efficiently Improve the Robustness of CNNs

Nikhil Kapoor, Chun Yuan, Jonas Löhdefink, Roland Zimmermann, Serin Varghese, Fabian Hüger, Nico M. Schmidt, Peter Schlicht, Tim Fingscheidt. A Self-Supervised Feature Map Augmentation (FMA) Loss and Combined Augmentations Finetuning to Efficiently Improve the Robustness of CNNs. In Björn Brücher, Oliver Wasenmüller, Mario Fritz, Hans-Joachim Hof, Christoph Krauß, editors, CSCS '20: Computer Science in Cars Symposium, Feldkirchen, Germany, December 2, 2020. ACM, 2020. [doi]

Abstract

Abstract is missing.