DiCo: Disentangled concept representation for text-to-image person re-identification

2026-01-01

Neurocomputing

AI Summary

This work addresses the significant semantic gap between visual and textual modalities and the challenge of fine-grained attribute alignment in text-to-image person re-identification. To this end, we propose DiCo, a novel framework that introduces, for the first time, a hierarchical structure of slots and concept blocks. Shared slots enable part-level cross-modal alignment, while each slot is decomposed into complementary concept blocks representing fine-grained attributes such as color, texture, and shape, thereby achieving disentangled representations. Integrating slot-based attention mechanisms with cross-modal contrastive learning, DiCo achieves state-of-the-art performance on CUHK-PEDES, ICFG-PEDES, and RSTPReid benchmarks, while also enabling high interpretability and fine-grained retrieval capabilities.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• 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
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Problem

• Text-to-image person re-identification
• modality gap
• fine-grained correspondence
• disentangled representation
• cross-modal alignment

Innovation

• disentangled representation
• slot-based modeling
• cross-modal alignment
• text-to-image person re-identification
• concept decomposition