Reconciling a Centroid-Hypothesis Conflict in Source-Free Domain Adaptation

TL;DR

This paper addresses the centroid-hypothesis conflict in source-free domain adaptation by aligning entropy minimization with pseudo-label cross-entropy, leading to improved adaptation performance across multiple datasets and architectures.

Contribution

It introduces a novel method to reconcile conflicting objectives in SFDA by aligning entropy and pseudo-label objectives, enhancing adaptation accuracy.

Findings

Achieves state-of-the-art results on three datasets.

Demonstrates consistency across different architectures.

Effectively reduces error accumulation in SFDA.

Abstract

Source-free domain adaptation (SFDA) aims to transfer knowledge learned from a source domain to an unlabeled target domain, where the source data is unavailable during adaptation. Existing approaches for SFDA focus on self-training usually including well-established entropy minimization techniques. One of the main challenges in SFDA is to reduce accumulation of errors caused by domain misalignment. A recent strategy successfully managed to reduce error accumulation by pseudo-labeling the target samples based on class-wise prototypes (centroids) generated by their clustering in the representation space. However, this strategy also creates cases for which the cross-entropy of a pseudo-label and the minimum entropy have a conflict in their objectives. We call this conflict the centroid-hypothesis conflict. We propose to reconcile this conflict by aligning the entropy minimization objective with that of the pseudo labels' cross entropy. We demonstrate the effectiveness of aligning the two loss objectives on three domain adaptation datasets. In addition, we provide state-of-the-art results using up-to-date architectures also showing the consistency of our method across these architectures.