Research map / Vision-language and multimodal / Human and embodied modeling
Generative Retrieval for Unsupervised Text-Based Person Search
Mang Ye, Yucheng Ji, Yang Bai, Min Cao, Siyuan Chai, Bo Du, Min Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Q&A base tier instead of template-based generation (ablation) | TierG · CUHK-PEDES | R@1: 47.27 -> 54.82 | Improved |
| use hard negatives (ISC) instead of random negatives (ablation) | TierG intermediate tier · CUHK-PEDES | R@1: 53.33 -> 54.58 | Improved |
| add stylized expansion instead of direct integration (ablation) | TierG advanced tier · CUHK-PEDES | R@1: 55.23 -> 55.58 | Improved |
| add intermediate and advanced tiers on top of Q&A base tier (ablation) | TierG · CUHK-PEDES | R@1: 54.82 -> 55.58 | Improved |
| add adaptive confidence weighting (ablation) | GTR+ retrieval · CUHK-PEDES | R@1: 55.58 -> 61.35 | Improved |
| use adaptive instead of fixed confidence weighting (ablation) | GTR+ retrieval · CUHK-PEDES | R@1: 58.72 -> 61.35 | Improved |
| replace similarity-based text token masking with adaptive weighting (ablation) | GTR+ retrieval · CUHK-PEDES | R@1: 55.60 -> 61.35 | Improved |
| add random image patch masking (ablation) | GTR+ retrieval · CUHK-PEDES | R@1: 55.58 -> 56.68 | Improved |
| increase hard negative pool from top-1 to top-10 (ablation) | TierG intermediate tier · CUHK-PEDES | R@1: 57.94 -> 59.50 | Improved |
| use K=2 Gaussian components instead of K=1 (ablation) | GMM confidence · CUHK-PEDES | R@1: 57.29 -> 61.35 | Improved |
| increase GMM components from K=2 to K=3 (ablation) | GMM confidence · CUHK-PEDES | R@1: 61.35 -> 59.31 | Hurt |
| pre-train on LargeFine-Person | GTR+ · CUHK-PEDES | R@1: 61.35 -> 64.65 | Improved |
Closest papers by meaning
- Exploring the Potential of Large Vision-Language Models for Unsupervised Text-Based Person Retrieval (AAAI 2025 · similarity 0.88)
- An Empirical Study of CLIP for Text-Based Person Search (AAAI 2024 · similarity 0.88)
- Automatic Synthetic Data and Fine-grained Adaptive Feature Alignment for Composed Person Retrieval (NeurIPS 2025 · similarity 0.88)
- UFineBench: Towards Text-based Person Retrieval with Ultra-fine Granularity (CVPR 2024 · similarity 0.88)
- Adaptive Uncertainty-Based Learning for Text-Based Person Retrieval (AAAI 2024 · similarity 0.87)
- Visual Perturbation for Text-Based Person Search (AAAI 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
