Research map / Label-efficient and robust learning / Recognition and long-tail
DC-Former: Diverse and Compact Transformer for Person Re-identification
Read the paperCheck 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 DC-Former instead of TransReID | ViT-B/16 person re-ID · MSMT17 | mAP: 67.4 -> 69.8 | Helped |
| Use DC-Former instead of AAformer | ViT-B/16 person re-ID · CUHK03-l | mAP: 77.8 -> 79.4 | Helped |
| Add a second class token without SDC (ablation) | Baseline · MSMT17 | mAP: 66.1 -> 66.9 | Helped |
| Add SDC to two class tokens (ablation) | Two-token model · MSMT17 | mAP: 66.9 -> 68.3 | Helped |
| Add SDC to two class tokens (ablation) | Two-token model · MSMT17 | Rank-1: 84.8 -> 85.5 | Helped |
| Concatenate two SDC-trained token embeddings instead of using one token (ablation) | DC-Former · MSMT17 | mAP: 67.9 -> 68.3 | Helped |
| Set SDC weight λ to 1 rather than 0 (ablation) | Two-token DC-Former · MSMT17 | mAP: 66.9 -> 68.3 | Helped |
| Increase λ beyond 1 (ablation) | Two-token DC-Former · MSMT17 | mAP | Hurt |
| Increase class-token count to N = 4 (ablation) | DC-Former · MSMT17 | mAP: 69.2% | Helped |
| Increase class-token count beyond the useful range (ablation) | DC-Former · MSMT17 | mAP | Hurt |
| Use DWC when token count is large (ablation) | DC-Former · MSMT17 | mAP: +1.0% | Helped |
| Use DWC when token count is small (less than 5) (ablation) | DC-Former · MSMT17 | mAP | No clear effect |
Closest papers by meaning
- Generalizable Person Re-Identification by Domain-Invariant Mapping Network (CVPR 2019 · similarity 0.88)
- Person30K: A Dual-Meta Generalization Network for Person Re-Identification (CVPR 2021 · similarity 0.87)
- From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · similarity 0.87)
- Resource Aware Person Re-Identification Across Multiple Resolutions (CVPR 2018 · similarity 0.87)
- Exploring the Camera Bias of Person Re-identification (ICLR 2025 · similarity 0.86)
- Instruct-ReID: A Multi-purpose Person Re-identification Task with Instructions (CVPR 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.