Research map / Architectures and efficiency / Attention and transformers
Transformer for Object Re-identification: A Survey
Mang Ye, Shuoyi Chen, Chenyue Li, Wei‐Shi Zheng, David Crandall, Bo Du
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 |
|---|---|---|---|
| replace CNN backbone with Transformer (TransReID vs AGW), supervised | supervised Re-ID · MSMT17 (supervised Re-ID) | mAP: 49.3 -> 69.4 | Improved |
| replace CNN backbone with Transformer (TransReID vs AGW), supervised | supervised Re-ID · Market-1501 (supervised Re-ID) | mAP: 87.8 -> 89.5 | Improved |
| pre-train Transformer with self-supervised learning on LUPerson instead of ImageNet | TransReID · MSMT17 (supervised Re-ID) | mAP: 69.4 -> 75.0 | Improved |
| pre-train Transformer with self-supervised learning on LUPerson instead of ImageNet | TransReID · Market-1501 (supervised Re-ID) | mAP: 89.5 -> 93.2 | Improved |
| swap CNN for Transformer with LUPerson SSL pre-training, unsupervised | Cluster-Contrast / PASS · MSMT17 (unsupervised Re-ID) | mAP: 31.4 -> 41.0 | Improved |
| LUPerson SSL pre-training instead of ImageNet supervised pre-training for a CNN | Cluster-Contrast · MSMT17 (unsupervised Re-ID) | mAP: 31.2 -> 31.4 | No clear effect |
| LUPerson SSL pre-training instead of ImageNet supervised pre-training for a CNN | Cluster-Contrast · Market-1501 (unsupervised Re-ID) | mAP: 83.0 -> 84.0 | Improved |
| use UntransReID baseline instead of TransReID-SSL* | TransReID-SSL* · MSMT17 (unsupervised Re-ID) | mAP: 48.2 -> 51.1 | Improved |
| use UntransReID baseline instead of TransReID-SSL* | TransReID-SSL* · Market-1501 (unsupervised Re-ID) | mAP: 89.9 -> 90.7 | Improved |
| add CLIP-based Transformer methods for text-image Re-ID | text-image Re-ID · CUHK-PEDES text-image Re-ID | Rank-1: 68.5 -> 76.8 | Improved |
| add CLIP-based Transformer methods for text-image Re-ID | text-image Re-ID · ICFG-PEDES text-image Re-ID | Rank-1: 60.1 -> 67.0 | Improved |
Closest papers by meaning
- TransReID: Transformer-Based Object Re-Identification (ICCV 2021 · similarity 0.93)
- Alice Benchmarks: Connecting Real World Re-Identification with the Synthetic (ICLR 2024 · similarity 0.89)
- Large-Scale Training Data Search for Object Re-Identification (CVPR 2023 · similarity 0.88)
- Magic Tokens: Select Diverse Tokens for Multi-modal Object Re-Identification (CVPR 2024 · similarity 0.87)
- MDReID: Modality-Decoupled Learning for Any-to-Any Multi-Modal Object Re-Identification (NeurIPS 2025 · similarity 0.86)
- TOP-ReID: Multi-Spectral Object Re-identification with Token Permutation (AAAI 2024 · similarity 0.86)
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.
