SOTAVerified

AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty

2019-12-05ICLR 2020Code Available1· sign in to hype

Dan Hendrycks, Norman Mu, Ekin D. Cubuk, Barret Zoph, Justin Gilmer, Balaji Lakshminarayanan

arXivPDF

Code Available — Be the first to reproduce this paper.

Reproduce

Code

Abstract

Modern deep neural networks can achieve high accuracy when the training distribution and test distribution are identically distributed, but this assumption is frequently violated in practice. When the train and test distributions are mismatched, accuracy can plummet. Currently there are few techniques that improve robustness to unforeseen data shifts encountered during deployment. In this work, we propose a technique to improve the robustness and uncertainty estimates of image classifiers. We propose AugMix, a data processing technique that is simple to implement, adds limited computational overhead, and helps models withstand unforeseen corruptions. AugMix significantly improves robustness and uncertainty measures on challenging image classification benchmarks, closing the gap between previous methods and the best possible performance in some cases by more than half.

Tasks

Domain Generalizationimage-classificationImage ClassificationOut-of-Distribution GeneralizationRobust Object Detection

Benchmark Results

DatasetModelMetricClaimedVerifiedStatus
ImageNet-CAugMix (ResNet-50)mean Corruption Error (mCE)65.3—Unverified
ImageNet-RAugMix (ResNet-50)Top-1 Error Rate58.9—Unverified
VizWiz-ClassificationResNet-50 (augmix)Accuracy - All Images42.2—Unverified

Reproductions