Domain Consistency Representation Learning for Lifelong Person Re-Identification

2024-09-30

AI Summary

To address the challenge of simultaneously preserving intra-domain discriminability and inter-domain consistency in lifelong person re-identification (LReID), this paper proposes a unified multi-granularity representation framework. Methodologically, it enhances fine-grained individual discrimination—e.g., clothing and accessories—via attribute-decoupled modeling and attribute-aware attention. Crucially, it introduces Domain-Consistent Representation (DCR) learning, a novel paradigm integrating Attribute-guided Forgetting mitigation (AF) and Progressive Knowledge Consolidation (KC) to alleviate catastrophic forgetting while ensuring stability across sequential domains. The framework synergistically combines contrastive learning, knowledge distillation, and disentangled representation learning. Extensive experiments demonstrate significant improvements in both discriminability and robustness under continual learning settings. Our approach achieves state-of-the-art performance on multiple LReID benchmarks. The source code is publicly available.

Technology Category

• Machine Learning: Life-Long and Continual Learning
• Computer Vision: Representation Learning for Vision
• Knowledge Representation and Reasoning: Argumentation

Application Category

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

Abstract

Lifelong person re-identification (LReID) exhibits a contradictory relationship between intra-domain discrimination and inter-domain gaps when learning from continuous data. Intra-domain discrimination focuses on individual nuances (i.e., clothing type, accessories, etc.), while inter-domain gaps emphasize domain consistency. Achieving a trade-off between maximizing intra-domain discrimination and minimizing inter-domain gaps is a crucial challenge for improving LReID performance. Most existing methods strive to reduce inter-domain gaps through knowledge distillation to maintain domain consistency. However, they often ignore intra-domain discrimination. To address this challenge, we propose a novel domain consistency representation learning (DCR) model that explores global and attribute-wise representations as a bridge to balance intra-domain discrimination and inter-domain gaps. At the intra-domain level, we explore the complementary relationship between global and attribute-wise representations to improve discrimination among similar identities. Excessive learning intra-domain discrimination can lead to catastrophic forgetting. We further develop an attribute-oriented anti-forgetting (AF) strategy that explores attribute-wise representations to enhance inter-domain consistency, and propose a knowledge consolidation (KC) strategy to facilitate knowledge transfer. Extensive experiments show that our DCR model achieves superior performance compared to state-of-the-art LReID methods. Our code is publicly available at https://github.com/LiuShiBen/DCR.

Problem

• Balancing intra-domain discrimination and inter-domain gaps in LReID
• Preventing catastrophic forgetting while learning intra-domain discrimination
• Enhancing domain consistency through global and attribute-wise representations

Innovation

• Balances intra-domain discrimination and inter-domain gaps
• Uses global and attribute-wise representations
• Implements attribute-oriented anti-forgetting strategy