AIDA-ReID: Adaptive Intermediate Domain Adaptation for Generalizable and Source-Free Person Re-Identification

2026-04-30

AI Summary

This work addresses the performance degradation of person re-identification in cross-domain deployment caused by variations in illumination, background, camera characteristics, and population distribution—particularly under source-free and multi-source scenarios where existing methods exhibit limited adaptability. To this end, we propose AIDA, an Adaptive Intermediate Domain Adaptation framework that introduces a novel dynamically controlled intermediate domain generation mechanism. By leveraging model uncertainty and training stability as feedback signals, AIDA adaptively adjusts feature mixing strategies and regularization strength. It further incorporates pseudo-mirror regularization and source-agnostic training techniques to enable effective multi-source domain generalization without requiring access to source data. Extensive experiments demonstrate that AIDA significantly outperforms state-of-the-art methods on multiple source-free domain generalization benchmarks, exhibiting superior generalization capability and robustness.

Abstract

Person re-identification (Re-ID) aims to match images of the same individual across non-overlapping camera views and remains challenging due to domain shifts caused by variations in illumination, background, camera characteristics, and population distributions. Although supervised models perform well under matched training and testing conditions, their performance degrades significantly when deployed in unseen environments. Existing intermediate domain approaches such as IDM and IDM++ alleviate this gap by constructing bridge feature distributions between domains; however, they rely on fixed mixing strategies and joint source-target access, limiting their applicability to multi-source and source-free settings. To address these limitations, this paper proposes Adaptive Intermediate Domain Adaptation (AIDA), also referred to as Source-Free Multi-Source Intermediate Domain Adaptation (SF-MIDA). The proposed framework treats intermediate-domain learning as a dynamically regulated process, where feature mixing and regularization strength are adaptively controlled using feedback signals derived from model uncertainty and training stability. A multi-source intermediate domain generator synthesizes diverse intermediate representations, while a pseudo-mirror regularization strategy preserves identity consistency under domain perturbations. Extensive experiments across domain generalization and source-free settings demonstrate the effectiveness of the proposed framework.

Problem

• person re-identification
• domain shift
• source-free
• domain generalization
• intermediate domain adaptation

Innovation

• Adaptive Intermediate Domain Adaptation
• Source-Free Re-Identification
• Domain Generalization
• Pseudo-Mirror Regularization
• Multi-Source Intermediate Domain