Reconciling a Centroid-Hypothesis Conflict in Source-Free Domain Adaptation
This work addresses a specific bottleneck in domain adaptation for scenarios where source data is unavailable, offering an incremental improvement over existing methods.
The paper tackles the problem of error accumulation in source-free domain adaptation by addressing a conflict between pseudo-labeling based on class centroids and entropy minimization, achieving state-of-the-art results on three datasets with consistent performance across architectures.
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.