Multi-level Consistency Learning for Semi-supervised Domain Adaptation

TL;DR

This paper introduces a Multi-level Consistency Learning framework for semi-supervised domain adaptation, aligning source and target domains at multiple levels to improve classification accuracy, achieving state-of-the-art results on benchmark datasets.

Contribution

The paper proposes a novel multi-level consistency learning framework that combines inter-domain alignment, intra-domain contrastive clustering, and sample-level self-training for SSDA.

Findings

Achieves state-of-the-art performance on VisDA2017, DomainNet, and Office-Home datasets.

Effectively aligns source and target domains using prototype-based optimal transport.

Improves target feature representations with class-wise contrastive clustering.

Abstract

Semi-supervised domain adaptation (SSDA) aims to apply knowledge learned from a fully labeled source domain to a scarcely labeled target domain. In this paper, we propose a Multi-level Consistency Learning (MCL) framework for SSDA. Specifically, our MCL regularizes the consistency of different views of target domain samples at three levels: (i) at inter-domain level, we robustly and accurately align the source and target domains using a prototype-based optimal transport method that utilizes the pros and cons of different views of target samples; (ii) at intra-domain level, we facilitate the learning of both discriminative and compact target feature representations by proposing a novel class-wise contrastive clustering loss; (iii) at sample level, we follow standard practice and improve the prediction accuracy by conducting a consistency-based self-training. Empirically, we verified the effectiveness of our MCL framework on three popular SSDA benchmarks, i.e., VisDA2017, DomainNet, and Office-Home datasets, and the experimental results demonstrate that our MCL framework achieves the state-of-the-art performance.