Exploiting the Intrinsic Neighborhood Structure for Source-free Domain Adaptation

TL;DR

This paper introduces a novel approach for source-free domain adaptation that leverages the intrinsic local neighborhood structure of target data to improve adaptation performance without access to source data.

Contribution

The method exploits local affinity and reciprocal neighbors in target data to enhance domain adaptation, achieving state-of-the-art results without source data access.

Findings

Effective use of local affinity improves adaptation.

Reciprocal neighbors reduce noise impact.

Expanded neighborhoods enhance feature aggregation.

Abstract

Domain adaptation (DA) aims to alleviate the domain shift between source domain and target domain. Most DA methods require access to the source data, but often that is not possible (e.g. due to data privacy or intellectual property). In this paper, we address the challenging source-free domain adaptation (SFDA) problem, where the source pretrained model is adapted to the target domain in the absence of source data. Our method is based on the observation that target data, which might no longer align with the source domain classifier, still forms clear clusters. We capture this intrinsic structure by defining local affinity of the target data, and encourage label consistency among data with high local affinity. We observe that higher affinity should be assigned to reciprocal neighbors, and propose a self regularization loss to decrease the negative impact of noisy neighbors. Furthermore, to aggregate information with more context, we consider expanded neighborhoods with small affinity values. In the experimental results we verify that the inherent structure of the target features is an important source of information for domain adaptation. We demonstrate that this local structure can be efficiently captured by considering the local neighbors, the reciprocal neighbors, and the expanded neighborhood. Finally, we achieve state-of-the-art performance on several 2D image and 3D point cloud recognition datasets. Code is available in https://github.com/Albert0147/SFDA_neighbors.