AETTA: Label-Free Accuracy Estimation for Test-Time Adaptation

TL;DR

AETTA introduces a label-free method for estimating the accuracy of test-time adaptation models using prediction disagreement, improving reliability in dynamic, unlabeled test scenarios.

Contribution

It proposes a novel accuracy estimation algorithm for TTA that does not require labels or re-training, enhancing adaptation reliability.

Findings

AETTA outperforms baselines with 19.8% more accurate estimates.

Demonstrates effectiveness in model recovery scenarios.

Validated across multiple TTA methods and baselines.

Abstract

Test-time adaptation (TTA) has emerged as a viable solution to adapt pre-trained models to domain shifts using unlabeled test data. However, TTA faces challenges of adaptation failures due to its reliance on blind adaptation to unknown test samples in dynamic scenarios. Traditional methods for out-of-distribution performance estimation are limited by unrealistic assumptions in the TTA context, such as requiring labeled data or re-training models. To address this issue, we propose AETTA, a label-free accuracy estimation algorithm for TTA. We propose the prediction disagreement as the accuracy estimate, calculated by comparing the target model prediction with dropout inferences. We then improve the prediction disagreement to extend the applicability of AETTA under adaptation failures. Our extensive evaluation with four baselines and six TTA methods demonstrates that AETTA shows an average of 19.8%p more accurate estimation compared with the baselines. We further demonstrate the effectiveness of accuracy estimation with a model recovery case study, showcasing the practicality of our model recovery based on accuracy estimation. The source code is available at https://github.com/taeckyung/AETTA.