Evidential Graph Contrastive Alignment for Source-Free Blending-Target Domain Adaptation

TL;DR

This paper introduces Evidential Contrastive Alignment (ECA), a novel method for source-free blending-target domain adaptation that improves pseudo-label quality and reduces distribution gaps across mixed target domains without source data.

Contribution

The paper proposes ECA, a new approach combining evidential learning and graph contrastive learning to handle label shifts and noisy pseudo labels in source-free blended target domain adaptation.

Findings

ECA outperforms existing methods on three standard DA datasets.

ECA achieves results comparable to methods using source data or domain labels.

ECA effectively reduces distribution gaps in blended target domains.

Abstract

In this paper, we firstly tackle a more realistic Domain Adaptation (DA) setting: Source-Free Blending-Target Domain Adaptation (SF-BTDA), where we can not access to source domain data while facing mixed multiple target domains without any domain labels in prior. Compared to existing DA scenarios, SF-BTDA generally faces the co-existence of different label shifts in different targets, along with noisy target pseudo labels generated from the source model. In this paper, we propose a new method called Evidential Contrastive Alignment (ECA) to decouple the blending target domain and alleviate the effect from noisy target pseudo labels. First, to improve the quality of pseudo target labels, we propose a calibrated evidential learning module to iteratively improve both the accuracy and certainty of the resulting model and adaptively generate high-quality pseudo target labels. Second, we design a graph contrastive learning with the domain distance matrix and confidence-uncertainty criterion, to minimize the distribution gap of samples of a same class in the blended target domains, which alleviates the co-existence of different label shifts in blended targets. We conduct a new benchmark based on three standard DA datasets and ECA outperforms other methods with considerable gains and achieves comparable results compared with those that have domain labels or source data in prior.