pSTarC: Pseudo Source Guided Target Clustering for Fully Test-Time Adaptation

TL;DR

pSTarC introduces a novel test-time adaptation method that leverages pseudo-source guided clustering to improve model performance under domain shifts without requiring source data, validated across multiple datasets.

Contribution

The paper proposes pSTarC, a new TTA approach that uses pseudo-source samples for clustering, enhancing adaptation efficiency and effectiveness under real-world domain shifts.

Findings

Significant accuracy improvements on VisDA, Office-Home, DomainNet-126, CIFAR-100C datasets.

Operates solely within test-time, no source data needed.

Effective for continuous TTA scenarios.

Abstract

Test Time Adaptation (TTA) is a pivotal concept in machine learning, enabling models to perform well in real-world scenarios, where test data distribution differs from training. In this work, we propose a novel approach called pseudo Source guided Target Clustering (pSTarC) addressing the relatively unexplored area of TTA under real-world domain shifts. This method draws inspiration from target clustering techniques and exploits the source classifier for generating pseudo-source samples. The test samples are strategically aligned with these pseudo-source samples, facilitating their clustering and thereby enhancing TTA performance. pSTarC operates solely within the fully test-time adaptation protocol, removing the need for actual source data. Experimental validation on a variety of domain shift datasets, namely VisDA, Office-Home, DomainNet-126, CIFAR-100C verifies pSTarC's effectiveness. This method exhibits significant improvements in prediction accuracy along with efficient computational requirements. Furthermore, we also demonstrate the universality of the pSTarC framework by showing its effectiveness for the continuous TTA framework. The source code for our method is available at https://manogna-s.github.io/pstarc