Generalizable Object Re-Identification via Visual In-Context Prompting

2025-08-28

AI Summary

Existing person re-identification (ReID) methods suffer from poor generalization, heavy reliance on large-scale category-specific annotations, and limited transferability to unseen categories. To address these limitations, we propose the Visual Context Prompting (VICP) framework—the first to integrate semantic rule reasoning from large language models (LLMs) into ReID. Given only a few positive and negative exemplars as contextual prompts, an LLM autonomously infers identity-discriminative rules and guides vision foundation models (e.g., DINO) to generate dynamic visual prompts, thereby achieving semantic–visual alignment. VICP requires no fine-tuning, parameter updates, or retraining, enabling zero-shot cross-category transfer. Evaluated on ShopID10K and multiple standard ReID benchmarks, VICP significantly outperforms state-of-the-art methods—particularly on unseen categories—demonstrating strong generalization capability and practical deployability.

Technology Category

• Computer Vision: Large Vision Models
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Prompt Engineering / Prompting

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Current object re-identification (ReID) methods train domain-specific models (e.g., for persons or vehicles), which lack generalization and demand costly labeled data for new categories. While self-supervised learning reduces annotation needs by learning instance-wise invariance, it struggles to capture extit{identity-sensitive} features critical for ReID. This paper proposes Visual In-Context Prompting~(VICP), a novel framework where models trained on seen categories can directly generalize to unseen novel categories using only extit{in-context examples} as prompts, without requiring parameter adaptation. VICP synergizes LLMs and vision foundation models~(VFM): LLMs infer semantic identity rules from few-shot positive/negative pairs through task-specific prompting, which then guides a VFM (eg, DINO) to extract ID-discriminative features via extit{dynamic visual prompts}. By aligning LLM-derived semantic concepts with the VFM's pre-trained prior, VICP enables generalization to novel categories, eliminating the need for dataset-specific retraining. To support evaluation, we introduce ShopID10K, a dataset of 10K object instances from e-commerce platforms, featuring multi-view images and cross-domain testing. Experiments on ShopID10K and diverse ReID benchmarks demonstrate that VICP outperforms baselines by a clear margin on unseen categories. Code is available at https://github.com/Hzzone/VICP.

Problem

• Generalizing object re-identification to unseen categories without retraining
• Reducing annotation dependency while maintaining identity-sensitive feature learning
• Aligning semantic concepts from LLMs with visual features from foundation models

Innovation

• Uses in-context examples as prompts for generalization
• Combines LLMs and vision models via dynamic visual prompts
• Aligns semantic concepts with pre-trained visual features