Benchmarking Neural Network Robustness to Common Corruptions and Perturbations

Code

• github.com/hendrycks/robustness
  OfficialIn paperpytorch★ 1,139
• github.com/mr-eggplant/sar
  pytorch★ 208
• github.com/mr-eggplant/eata
  pytorch★ 138
• github.com/YutingLi0606/SURE
  pytorch★ 75
• github.com/facebookresearch/augmentation-corruption
  pytorch★ 46
• github.com/automl/nes
  pytorch★ 33
• github.com/allenai/robustnav
  pytorch★ 31
• github.com/yaodongyu/projnorm
  pytorch★ 19
• github.com/deeplearning-wisc/hypo
  pytorch★ 18
• github.com/MKYucel/zero_shot_corruption_benchmarks
  pytorch★ 5

Abstract

In this paper we establish rigorous benchmarks for image classifier robustness. Our first benchmark, ImageNet-C, standardizes and expands the corruption robustness topic, while showing which classifiers are preferable in safety-critical applications. Then we propose a new dataset called ImageNet-P which enables researchers to benchmark a classifier's robustness to common perturbations. Unlike recent robustness research, this benchmark evaluates performance on common corruptions and perturbations not worst-case adversarial perturbations. We find that there are negligible changes in relative corruption robustness from AlexNet classifiers to ResNet classifiers. Afterward we discover ways to enhance corruption and perturbation robustness. We even find that a bypassed adversarial defense provides substantial common perturbation robustness. Together our benchmarks may aid future work toward networks that robustly generalize.

Tasks

Benchmark Results

Reproductions