AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty (2019-12-05T00:00:00.000000Z)

TL;DR

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.

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.

Authors

E. D. Cubuk

9 papers

J. Gilmer

7 papers

Balaji Lakshminarayanan

8 papers

TL;DR

Abstract

Authors

References49 items

An Empirical Evaluation on Robustness and Uncertainty of Regularization Methods

Randaugment: Practical automated data augmentation with a reduced search space

A Discussion of 'Adversarial Examples Are Not Bugs, They Are Features': Adversarial Example Researchers Need to Expand What is Meant by 'Robustness'

A Fourier Perspective on Model Robustness in Computer Vision

Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift

Improving Robustness Without Sacrificing Accuracy with Patch Gaussian Augmentation

Adversarial Training Can Hurt Generalization

CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features

Unsupervised Data Augmentation

Making Convolutional Networks Shift-Invariant Again

Using Videos to Evaluate Image Model Robustness

Benchmarking Neural Network Robustness to Common Corruptions and Perturbations

Using Pre-Training Can Improve Model Robustness and Uncertainty

Data Augmentation Using Random Image Cropping and Patching for Deep CNNs

Deep Anomaly Detection with Outlier Exposure

MixUp as Locally Linear Out-Of-Manifold Regularization

Generalisation in humans and deep neural networks

Evaluating and Understanding the Robustness of Adversarial Logit Pairing

Motivating the Rules of the Game for Adversarial Example Research

Why do deep convolutional networks generalize so poorly to small image transformations?

AutoAugment: Learning Augmentation Policies from Data

Adversarial Logit Pairing

Detecting and Correcting for Label Shift with Black Box Predictors

Between-Class Learning for Image Classification

mixup: Beyond Empirical Risk Minimization

Random Erasing Data Augmentation

Improved Regularization of Convolutional Neural Networks with Cutout

Towards Deep Learning Models Resistant to Adversarial Attacks

On Calibration of Modern Neural Networks

Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour

Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles

Examining the Impact of Blur on Recognition by Convolutional Networks

Aggregated Residual Transformations for Deep Neural Networks

A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks

Densely Connected Convolutional Networks

SGDR: Stochastic Gradient Descent with Warm Restarts

Wide Residual Networks

Improving the Robustness of Deep Neural Networks via Stability Training

Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks

Deep Residual Learning for Image Recognition

Posterior calibration and exploratory analysis for natural language processing models

Striving for Simplicity: The All Convolutional Net

Learning with Pseudo-Ensembles

ImageNet classification with deep convolutional neural networks

The conference paper

Unbiased look at dataset bias

ImageNet: A large-scale hierarchical image database

Testing Robustness Against Unforeseen Adversaries

Learning Multiple Layers of Features from Tiny Images

Field of Study

Journal Information

Name

Volume

Venue Information

Name

Type

URL

Alternate Names