1 paper · 1 filter
Alexander Bartler, Florian Bender, Felix Wiewel +1
Nowadays, deep neural networks outperform humans in many tasks. However, if the input distribution drifts away from the one used in training, their performance drops significantly.…