Abstract:Source-free Unsupervised Domain Adaptation (SFDA) aims to classify targetsamples by only accessing a pre-trained source model and unlabelled targetsamples. Since no source data is available, transferring the knowledge from thesource domain to the target domain is challenging. Existing methods normallyexploit the pair-wise relation among target samples and attempt to discovertheir correlations by clustering these samples based on semantic features. Thedrawback of these methods includes: 1) the pair-wise relation is limited toexposing the underlying correlations of two more samples, hindering theexploration of the structural information embedded in the target domain; 2) theclustering process only relies on the semantic feature, while overlooking thecritical effect of domain shift, i.e., the distribution differences between thesource and target domains. To address these issues, we propose a new SFDAmethod that exploits the high-order neighborhood relation and explicitly takesthe domain shift effect into account. Specifically, we formulate the SFDA as aHypergraph learning problem and construct hyperedges to explore the local groupand context information among multiple samples. Moreover, we integrate aself-loop strategy into the constructed hypergraph to elegantly introduce thedomain uncertainty of each sample. By clustering these samples based onhyperedges, both the semantic feature and domain shift effects are considered.We then describe an adaptive relation-based objective to tune the model withsoft attention levels for all samples. Extensive experiments are conducted onOffice-31, Office-Home, VisDA, and PointDA-10 datasets. The results demonstratethe superiority of our method over state-of-the-art counterparts.

High-order Neighborhoods Know More: HyperGraph Learning Meets Source-free Unsupervised Domain Adaptation

A Framework for Self-Supervised Federated Domain Adaptation

Target-Discriminability-Induced Multi-Source-Free Domain Adaptation

Exploiting the Intrinsic Neighborhood Structure for Source-free Domain Adaptation

Attracting and Dispersing: A Simple Approach for Source-free Domain Adaptation

Unified multi-level neighbor clustering for Source-Free Unsupervised Domain Adaptation

Hierarchical Unsupervised Relation Distillation for Source Free Domain Adaptation

Trust Your Good Friends: Source-Free Domain Adaptation by Reciprocal Neighborhood Clustering

Source-Free Domain Adaptation Via Distribution Estimation

A Comprehensive Survey on Source-Free Domain Adaptation

What Has Been Overlooked in Contrastive Source-Free Domain Adaptation: Leveraging Source-Informed Latent Augmentation within Neighborhood Context

SF(DA)$^2$: Source-free Domain Adaptation Through the Lens of Data Augmentation

Unified Source-Free Domain Adaptation

Unveiling the Superior Paradigm: A Comparative Study of Source-Free Domain Adaptation and Unsupervised Domain Adaptation

Source-Free Domain Adaptation Via Target Prediction Distribution Searching

Source-Free Domain Adaptation for Semantic Segmentation

Generation, Augmentation, and Alignment: A Pseudo-Source Domain Based Method for Source-Free Domain Adaptation.

Self-Mining the Confident Prototypes for Source-Free Unsupervised Domain Adaptation in Image Segmentation

Upcycling Models under Domain and Category Shift

UC-SFDA: Source-free Domain Adaptation Via Uncertainty Prediction and Evidence-Based Contrastive Learning

GLC++: Source-Free Universal Domain Adaptation through Global-Local Clustering and Contrastive Affinity Learning