SoTTA: Robust Test-Time Adaptation on Noisy Data Streams

TL;DR

SoTTA introduces a robust test-time adaptation method that effectively filters noisy data and stabilizes model parameters, significantly improving performance in noisy data streams compared to existing approaches.

Contribution

The paper proposes SoTTA, a novel TTA algorithm that enhances robustness to noisy test samples through high-confidence sampling and entropy-sharpness minimization.

Findings

Outperforms state-of-the-art TTA methods in noisy scenarios.

Maintains comparable accuracy to existing methods without noise.

Effectively filters out noisy samples during adaptation.

Abstract

Test-time adaptation (TTA) aims to address distributional shifts between training and testing data using only unlabeled test data streams for continual model adaptation. However, most TTA methods assume benign test streams, while test samples could be unexpectedly diverse in the wild. For instance, an unseen object or noise could appear in autonomous driving. This leads to a new threat to existing TTA algorithms; we found that prior TTA algorithms suffer from those noisy test samples as they blindly adapt to incoming samples. To address this problem, we present Screening-out Test-Time Adaptation (SoTTA), a novel TTA algorithm that is robust to noisy samples. The key enabler of SoTTA is two-fold: (i) input-wise robustness via high-confidence uniform-class sampling that effectively filters out the impact of noisy samples and (ii) parameter-wise robustness via entropy-sharpness minimization that improves the robustness of model parameters against large gradients from noisy samples. Our evaluation with standard TTA benchmarks with various noisy scenarios shows that our method outperforms state-of-the-art TTA methods under the presence of noisy samples and achieves comparable accuracy to those methods without noisy samples. The source code is available at https://github.com/taeckyung/SoTTA .