Research map / Vision-language and multimodal / Human and embodied modeling
PLOT: Text-based Person Search with Part Slot Attention for Corresponding Part Discovery
Jicheol Park, Dongwon Kim, Boseung Jeong, Suha Kwak
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 part embeddings (PartNCE) on top of global alignment with CMLM (ablation) | PLOT · CUHK-PEDES | R@1: 72.65 -> 74.85 | Improved |
| add PartID loss (ablation) | PLOT · CUHK-PEDES | R@1: 74.85 -> 75.28 | Improved |
| add CMLM loss (ablation) | PLOT global-only · CUHK-PEDES | R@1: 71.83 -> 72.65 | Improved |
| add identity loss to global alignment (ablation) | global-only baseline · CUHK-PEDES | R@10: 92.74 -> 92.58 | No clear effect |
| use slot-attention part discovery instead of TIPCB heuristic cropping (ablation) | PLOT · CUHK-PEDES | R@1: 73.23 -> 75.28 | Improved |
| use slot attention instead of PAT cross-attention queries (ablation) | PLOT · CUHK-PEDES | R@1: 72.76 -> 75.28 | Improved |
| add text-based dynamic part attention (TDPA) instead of cumulative sum (ablation) | PLOT · CUHK-PEDES | R@1: 73.96 -> 75.28 | Improved |
| add GRU update in PSA block (ablation) | PLOT · CUHK-PEDES | R@1: 74.33 -> 75.28 | Improved |
| increase part slots from 8 to 10 (ablation) | PLOT · CUHK-PEDES | R@1: 75.28 -> 74.51 | No clear effect |
| reduce part slots from 8 to 4 (ablation) | PLOT · CUHK-PEDES | R@1: 75.28 -> 73.79 | Hurt |
| use PLOT instead of RaSa re-ranking | text-based person search · CUHK-PEDES | inference time (s): 1168 -> 16 | Improved |
| PLOT versus RaSa on RSTPReid | PLOT · RSTPReid | R@1: 66.90 -> 61.80 | Hurt |
Closest papers by meaning
- A Survey on Text-Based Person Search (TPAMI 2026 · similarity 0.87)
- Quota-Calibrated Fine-Grained Alignment with Context-Aware Marginals for Text-based Person Retrieval (CVPR 2026 · similarity 0.85)
- Automatic Synthetic Data and Fine-grained Adaptive Feature Alignment for Composed Person Retrieval (NeurIPS 2025 · similarity 0.85)
- Diverse Person: Customize Your Own Dataset for Text-Based Person Search (AAAI 2024 · similarity 0.83)
- Visual Perturbation for Text-Based Person Search (AAAI 2025 · similarity 0.83)
- An Empirical Study of CLIP for Text-Based Person Search (AAAI 2024 · similarity 0.83)
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.
