Papers › Measuring Robustness to Natural Distribution Shifts in Image Classification

Measuring Robustness to Natural Distribution Shifts in Image Classification

1 Jul 2020NeurIPS 2020 12arXiv:2007.00644archive 2025-07-28

Rohan Taori, Achal Dave, Vaishaal Shankar, Nicholas Carlini, Benjamin Recht, Ludwig Schmidt

We study how robust current ImageNet models are to distribution shifts arising from natural variations in datasets. Most research on robustness focuses on synthetic image perturbations (noise, simulated weather artifacts, adversarial examples, etc.), which leaves open how robustness on synthetic distribution shift relates to distribution shift arising in real data. Informed by an evaluation of 204 ImageNet models in 213 different test conditions, we find that there is often little to no transfer of robustness from current synthetic to natural distribution shift. Moreover, most current techniques provide no robustness to the natural distribution shifts in our testbed. The main exception is training on larger and more diverse datasets, which in multiple cases increases robustness, but is still far from closing the performance gaps. Our results indicate that distribution shifts arising in real data are currently an open research problem. We provide our testbed and data as a resource for future work at https://modestyachts.github.io/imagenet-testbed/ .

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

modestyachts/imagenet-testbed officialmentioned on GitHubpytorchMIT report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

ClassificationDomain GeneralizationGeneral ClassificationImage Classificationimage-classification

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Domain Generalization VizWiz-Classification ResNet-50 (IN-C) Accuracy - All Images 38.8 #41 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C) Accuracy - Clean Images 42.9 #41 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C) Accuracy - Corrupted Images 33.6 #41 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_brightness) Accuracy - All Images 38.8 #42 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_brightness) Accuracy - Clean Images 43.5 #42 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_brightness) Accuracy - Corrupted Images 32.5 #42 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_spatter) Accuracy - All Images 38.3 #48 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_spatter) Accuracy - Clean Images 42.7 #48 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_spatter) Accuracy - Corrupted Images 31.4 #48 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_saturate) Accuracy - All Images 38.2 #50 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_saturate) Accuracy - Clean Images 42.4 #50 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_saturate) Accuracy - Corrupted Images 32.4 #50 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_pixelate) Accuracy - All Images 37.4 #52 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_pixelate) Accuracy - Clean Images 41.4 #52 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_pixelate) Accuracy - Corrupted Images 30.9 #52 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_contrast) Accuracy - All Images 36.5 #59 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_contrast) Accuracy - Clean Images 40.9 #59 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_contrast) Accuracy - Corrupted Images 30.7 #59 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_jpeg_compression) Accuracy - All Images 36.5 #60 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_jpeg_compression) Accuracy - Clean Images 41.3 #60 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_jpeg_compression) Accuracy - Corrupted Images 30.3 #60 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_gaussian_noise) Accuracy - All Images 36.4 #61 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_gaussian_noise) Accuracy - Clean Images 40.6 #61 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_gaussian_noise) Accuracy - Corrupted Images 30.2 #61 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_frost) Accuracy - All Images 36.1 #63 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_frost) Accuracy - Clean Images 40 #63 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_frost) Accuracy - Corrupted Images 29.7 #63 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_fog_aws) Accuracy - All Images 35.9 #65 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_fog_aws) Accuracy - Clean Images 39.9 #65 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_fog_aws) Accuracy - Corrupted Images 30.3 #65 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_motion_blur) Accuracy - All Images 35.7 #67 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_motion_blur) Accuracy - Clean Images 39.6 #67 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_motion_blur) Accuracy - Corrupted Images 30.2 #67 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_zoom_blur) Accuracy - All Images 32.7 #81 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_zoom_blur) Accuracy - Clean Images 36.6 #81 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_zoom_blur) Accuracy - Corrupted Images 28.3 #81 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_greyscale) Accuracy - All Images 30.2 #84 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_greyscale) Accuracy - Clean Images 34.3 #84 of 90 Archive leaderboard report
Domain Generalization VizWiz-Classification ResNet-50 (IN-C_greyscale) Accuracy - Corrupted Images 24.3 #84 of 90 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections