# hendrycks/robustness

Corruption and Perturbation Robustness (ICLR 2019)

Repository: https://github.com/hendrycks/robustness
Canonical: https://ross.abutalabs.com/products/robustness
Language: Python
License: Apache-2.0
License Family: permissive
Topics: imagenet, robustness, pytorch, deep-learning, convolutional-neural-networks, computer-vision, machine-learning, domain-generalization, ml-safety
Last push: 2022-08-24T14:27:43+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3071, "days_push": 1470, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1172, forks 149 (observed 2026-08-28T04:03:51.706205+00:00)

## What it is
Datasets and evaluation code for benchmarking neural network robustness to common corruptions and perturbations, from the ICLR 2019 paper by Hendrycks and Dietterich. It provides ImageNet-C, Tiny ImageNet-C, and CIFAR-10/100-C corruption benchmarks plus a leaderboard of model robustness results.

## Use cases
- evaluate image classifier robustness to common corruptions
- download imagenet-c corrupted image datasets
- benchmark my resnet-50 on imagenet-c mean corruption error
- test model robustness to gaussian noise and blur perturbations
- compare corruption robustness of cnn training methods
- run cifar-10-c robustness evaluation in pytorch

## When to choose
- you need standardized corruption benchmarks like ImageNet-C or CIFAR-10-C
- you want to compare your model's mCE against published robustness results
- you are doing ML safety or domain generalization research in PyTorch

## When to avoid
- you need adversarial (worst-case) attack robustness rather than common corruptions
- you work outside image classification or non-PyTorch frameworks
- you need actively maintained tooling rather than a research artifact

## Facets
- artifact type: dataset
- maturity: maintenance
- function: benchmarking, machine-learning, computer-vision, image-processing, testing
- domain: machine-learning, deep-learning, computer-vision, artificial-intelligence
- platform: python, windows
- tags: imagenet-c, robustness, corruptions, pytorch, benchmark-datasets, ml-safety, domain-generalization, iclr-2019, linux, macos

## Member repositories
- hendrycks/robustness (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:51.706205+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:28:27.131185+00:00, confidence not recorded.
  - readme: https://github.com/hendrycks/robustness (fetched 2026-08-28T04:03:51.706205+00:00, sha 56eee6f333e1)
- Data as of 2026-08-30T08:39:29.467469+00:00.