1 paper · 2 filters
Markus Heinonen, Ba-Hien Tran, Michael Kampffmeyer +1
Introducing training-time augmentations is a key technique to enhance generalization and prepare deep neural networks against test-time corruptions. Inspired by the success of gene…