Learning Loss for Test-Time Augmentation

TL;DR

This paper introduces a novel instance-level test-time augmentation method that predicts and applies the most suitable transformations for each input, enhancing neural network robustness against various corruptions.

Contribution

It proposes an auxiliary module to predict transformation losses and select optimal augmentations at test time, a new approach for adaptive test-time augmentation.

Findings

Improves robustness on image classification benchmarks

Outperforms traditional test-time augmentation methods

Enhances model performance against corruptions

Abstract

Data augmentation has been actively studied for robust neural networks. Most of the recent data augmentation methods focus on augmenting datasets during the training phase. At the testing phase, simple transformations are still widely used for test-time augmentation. This paper proposes a novel instance-level test-time augmentation that efficiently selects suitable transformations for a test input. Our proposed method involves an auxiliary module to predict the loss of each possible transformation given the input. Then, the transformations having lower predicted losses are applied to the input. The network obtains the results by averaging the prediction results of augmented inputs. Experimental results on several image classification benchmarks show that the proposed instance-aware test-time augmentation improves the model's robustness against various corruptions.