Dual-level Modality Debiasing Learning for Unsupervised Visible-Infrared Person Re-Identification

2025-12-03

AI Summary

To address modality bias introduced by unimodal pretraining in unsupervised visible-infrared person re-identification (USL-VI-ReID), this paper proposes a dual-level debiasing learning framework. At the model level, we design a causality-inspired adjustment intervention module to disentangle spurious modality-specific correlations. At the optimization level, we formulate a collaborative unbiased training strategy that blocks bias propagation across data, label, and feature levels. Our method integrates modality-specific augmentation, label refinement, and cross-modal feature alignment to learn modality-invariant representations with enhanced identity discriminability. Extensive experiments on standard benchmarks—including SYSU-MM01 and RegDB—demonstrate significant improvements in Rank-1 accuracy and mAP over prior methods, validating both effectiveness and generalizability of the proposed approach.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Two-stage learning pipeline has achieved promising results in unsupervised visible-infrared person re-identification (USL-VI-ReID). It first performs single-modality learning and then operates cross-modality learning to tackle the modality discrepancy. Although promising, this pipeline inevitably introduces modality bias: modality-specific cues learned in the single-modality training naturally propagate into the following cross-modality learning, impairing identity discrimination and generalization. To address this issue, we propose a Dual-level Modality Debiasing Learning (DMDL) framework that implements debiasing at both the model and optimization levels. At the model level, we propose a Causality-inspired Adjustment Intervention (CAI) module that replaces likelihood-based modeling with causal modeling, preventing modality-induced spurious patterns from being introduced, leading to a low-biased model. At the optimization level, a Collaborative Bias-free Training (CBT) strategy is introduced to interrupt the propagation of modality bias across data, labels, and features by integrating modality-specific augmentation, label refinement, and feature alignment. Extensive experiments on benchmark datasets demonstrate that DMDL could enable modality-invariant feature learning and a more generalized model.

Problem

• Addresses modality bias in unsupervised visible-infrared person re-identification
• Proposes dual-level debiasing at model and optimization stages
• Enhances modality-invariant feature learning for better generalization

Innovation

• Causality-inspired Adjustment Intervention for low-biased modeling
• Collaborative Bias-free Training with multi-level debiasing
• Dual-level framework addressing modality bias in ReID