Research map / Vision-language and multimodal / Cross-modal learning
Knowing Where to Focus: Attention-Guided Alignment for Text-based Person Search
Lei Tan, Weihao Li, Pingyang Dai, Jie Chen, Liujuan Cao, Rongrong Ji
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 Attention-Guided Mask (AGM) instead of random masking (ablation) | baseline · CUHK-PEDES | Rank-1 / mAP: 76.03 -> 77.58 R@1; 69.23 -> 70.50 mAP | Improved |
| add Text Enrichment Module (TEM) (ablation) | baseline · CUHK-PEDES | Rank-1: 76.03 -> 76.87 | Improved |
| add TEM on top of AGM (ablation) | AGM · CUHK-PEDES | Rank-1 / mAP: 77.58 -> 78.36 R@1; 70.50 -> 70.72 mAP | Improved |
| mask only semantically vacuous words (Picked) instead of no MLM (ablation) | baseline · CUHK-PEDES | Rank-1: 76.03 -> 75.50 | Hurt |
| random masking MLM vs no MLM (ablation) | baseline · CUHK-PEDES | Rank-1: 76.03 -> 76.51 | Improved |
| replace random masking with AGM (lower vacuous ratio) (ablation) | MLM · CUHK-PEDES | Rank-1: 76.51 -> 77.58 | Improved |
| use EMA aggregation of layer class attention instead of mean (ablation) | AGM · CUHK-PEDES | Rank-1: 78.18 -> 78.36 | Improved |
| use EMA aggregation instead of last-layer attention (ablation) | AGM · CUHK-PEDES | Rank-1: 78.08 -> 78.36 | Improved |
| vary TEM replacement probability p_tem from 0.1 to 0.4 (ablation) | TEM · CUHK-PEDES | Rank-1: 78.23 -> 78.39 -> 78.36 -> 78.11 | No clear effect |
| vary number of TEM candidate words k (3, 5, 10) (ablation) | TEM · CUHK-PEDES | Rank-1: 78.00 -> 78.36 -> 78.05 | Mixed |
| vary EMA beta (0.90, 0.95, 0.99) (ablation) | AGM · CUHK-PEDES | Rank-1: 78.21 -> 78.36 -> 77.95 | Mixed |
| raise lower bound alpha1 of masking probability from -0.05 to 0.1 (ablation) | AGM · CUHK-PEDES | Rank-1: 76.03 -> 75.50 -> 76.51 -> 77.58 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PLOT — PLOT: Prompt Learning with Optimal Transport for Vision-Language Models (ICLR 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- 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)
- Chat-based Person Retrieval via Dialogue-Refined Cross-Modal Alignment (CVPR 2025 · similarity 0.87)
- An Empirical Study of CLIP for Text-Based Person Search (AAAI 2024 · similarity 0.86)
- Text Embedding is Not All You Need: Attention Control for Text-to-Image Semantic Alignment with Text Self-Attention Maps (CVPR 2025 · similarity 0.85)
- Unifying Multi-Modal Uncertainty Modeling and Semantic Alignment for Text-to-Image Person Re-identification (AAAI 2024 · similarity 0.85)
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.
