Research map / Detection and segmentation / CNN features
PSTR: End-to-End One-Step Person Search With Transformers
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use PSTR with ResNet50 instead of AlignPS with ResNet50 | Person search on PRW · PRW test | mAP: 45.9 -> 49.5 | Helped |
| Use shared multi-level re-id decoding instead of single-level re-id decoding (ablation) | PSTR · PRW test | mAP: 47.4 -> 51.9 | Helped |
| Use shared multi-level re-id decoding instead of single-level re-id decoding (ablation) | PSTR · PRW test | Top-1: 84.9 -> 88.3 | Helped |
| Use shared re-id decoder instead of parallel re-id decoders (ablation) | PSTR · PRW test | mAP: 51.4 -> 51.9 | Helped |
| Use backbone P4 features rather than detection encoder features as re-id input (ablation) | PSTR · PRW test | mAP: 47.5 -> 51.9 | Helped |
| Use part attention instead of deformable attention (ablation) | PSTR re-id decoder · PRW test | mAP: 51.5 -> 51.9 | Helped |
| Use part attention instead of deformable attention (ablation) | PSTR re-id decoder · PRW test | Top-1: 87.8 -> 88.3 | Helped |
| Add P3 to the P4 re-id scale (ablation) | PSTR multi-scale re-id decoder · PRW test | mAP: 51.9 -> 54.7 | Helped |
| Add P2 to the P4+P3 re-id scales (ablation) | PSTR multi-scale re-id decoder · PRW test | mAP: 54.7 -> 56.5 | Helped |
| Use PSTR's person-search-specialized module instead of a shared standard encoder-decoder (ablation) | One-step transformer person search · PRW test | mAP: 23.1 -> 51.9 | Helped |
| Use PSTR's person-search-specialized module instead of separate detection and re-id encoder-decoders (ablation) | One-step transformer person search · PRW test | mAP: 44.5 -> 51.9 | Helped |
| Use the PVTv2-B2 backbone instead of ResNet50 | PSTR · PRW test | mAP: 49.5 -> 56.5 | Helped |
Closest papers by meaning
- Person Re-Identification in the Wild (CVPR 2017 · similarity 0.87)
- Hierarchical Online Instance Matching for Person Search (AAAI 2020 · similarity 0.87)
- Diverse Part Discovery: Occluded Person Re-Identification With Part-Aware Transformer (CVPR 2021 · similarity 0.87)
- Diverse Knowledge Distillation for End-to-End Person Search (AAAI 2021 · similarity 0.87)
- Attribute-Guided Pedestrian Retrieval: Bridging Person Re-ID with Internal Attribute Variability (CVPR 2024 · similarity 0.85)
- Cheb-GR: Rethinking K-nearest Neighbor Search in Re-ranking for Person Re-identification (CVPR 2025 · 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.