Research map / Label-efficient and robust learning / Recognition and long-tail
A Versatile Framework for Multi-Scene Person Re-Identification
Wei‐Shi Zheng, Junkai Yan, Yi-Xing Peng
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 |
|---|---|---|---|
| train VersReID prompt-based twin model vs best multi-scene baseline ReIDCaps | ViT-B multi-scene ReID · Multi-scene joint testing set (multi-scene training) | Rank-1 (%): 74.0 -> 76.3 | Improved |
| use ViT-B instead of ResNet-50 backbone for AIM | AIM · Multi-scene joint testing set (multi-scene training) | Rank-1 (%): 65.2 -> 68.7 | Improved |
| multi-scene joint training of VersReID instead of single-dataset ViT-B (MPDA) | VersReID · Seven-dataset average | Rank-1 (%): 76.3 -> 77.9 | Improved |
| multi-scene joint training instead of single-dataset training | VersReID · PRCC | Rank-1 (%): 54.4 -> 61.3 | Improved |
| multi-scene joint training instead of single-dataset training | VersReID · SYSU-mm01 | Rank-1 (%): 64.6 -> 67.7 | Improved |
| distill scene-label-free V-Branch from ReID Bank (ablation) | VersReID · Multi-scene joint testing set (multi-scene training) | Rank-1 (%): 78.7 -> 76.3 | Hurt |
| distill V-Branch from ReID Bank (ablation) | VersReID · MSMT17 | Rank-1 (%): 87.9 -> 88.5 | Improved |
| use overlapping patch embedding layer | VersReID · Celeb-ReID | Rank-1 (%): 60.8 -> 61.7 | Improved |
| use overlapping patch embedding layer | VersReID · Occ-Duke | Rank-1 (%): 72.9 -> 75.2 | Improved |
| use overlapping patch embedding layer | VersReID · Market-1501 | Rank-1 (%): 96.9 -> 96.8 | No clear effect |
| use overlapping patch embedding layer | VersReID · Joint testing set (computational cost) | Rank-1 (%) / GMAC: 16.52 -> 27.04 GMAC; 76.3 -> 77.2 Rank-1 | Mixed |
| initialize prompts with pre-trained CLIP text encoder instead of random (ablation) | scene-specific prompts · multi-scene ReID | Rank-1 (%): no improvement reported | No clear effect |
Closest papers by meaning
- Instruct-ReID: A Multi-purpose Person Re-identification Task with Instructions (CVPR 2024 · similarity 0.91)
- Instruct-ReID++: Towards Universal Purpose Instruction-Guided Person Re-Identification (TPAMI 2025 · similarity 0.91)
- Exploring the Camera Bias of Person Re-identification (ICLR 2025 · similarity 0.90)
- Learning Deep Context-Aware Features Over Body and Latent Parts for Person Re-Identification (CVPR 2017 · similarity 0.90)
- Multi-Prompts Learning with Cross-Modal Alignment for Attribute-Based Person Re-identification (AAAI 2024 · similarity 0.89)
- From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · similarity 0.89)
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.
