Research map / Label-efficient and robust learning / Recognition and long-tail
Densely Semantically Aligned Person Re-Identification
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 |
|---|---|---|---|
| add random erasing (ablation) | Baseline on Market1501 (SQ) · Market1501 (SQ) | mAP: 76.4 -> 78.6 | Helped |
| add label smoothing to the random-erasing baseline (ablation) | Baseline (RE) on Market1501 (SQ) · Market1501 (SQ) | mAP: +2.6% | Helped |
| use a two-branch head (ablation) | Baseline (RE+LS) on Market1501 (SQ) · Market1501 (SQ) | mAP: 81.2 -> 83.4 | Helped |
| add dense-semantic alignment guidance | Baseline (RE+LS) on Market1501 (SQ) · Market1501 (SQ) | mAP: +6.4% | Helped |
| add dense-semantic alignment guidance | Baseline (Two branches) on Market1501 (SQ) · Market1501 (SQ) | mAP: +4.2% | Helped |
| use only the global branch with semantic alignment (ablation) | Baseline (RE+LS) on Market1501 (SQ) · Market1501 (SQ) | mAP: +3.5% | Helped |
| use only the local branch with semantic alignment (ablation) | Baseline (RE+LS) on Market1501 (SQ) · Market1501 (SQ) | mAP: +2.0% | Helped |
| combine global and part-aware features (ablation) | DSA two-stream framework on Market1501 (SQ) · Market1501 (SQ) | mAP: +2.8% vs DSA-Global(Single); +4.3% vs DSA-Local(Single) | Helped |
| jointly train the streams (ablation) | DSA global branch on Market1501 (SQ) · Market1501 (SQ) | mAP: 84.7 -> 87.4 | Helped |
| replace coarse cropped parts with densely aligned DSAP-images (ablation) | CSA framework on Market1501 (SQ) · Market1501 (SQ) | mAP: +3.5% | Helped |
| replace concatenation plus fully connected projection with element-wise addition (ablation) | Two-stream fusion on Market1501 (SQ) · Market1501 (SQ) | mAP: 81.6 -> 87.6 | Helped |
| discard DSAG-Stream at inference while retaining its training guidance (ablation) | DSA-reID on Market1501 (SQ) · Market1501 (SQ) | mAP: 87.5 -> 87.6 | No clear effect |
Closest papers by meaning
- Semantics-Aligned Representation Learning for Person Re-Identification (AAAI 2020 · similarity 0.92)
- Human Semantic Parsing for Person Re-Identification (CVPR 2018 · similarity 0.90)
- Unsupervised Domain Adaptation for Person Re-identification via Heterogeneous Graph Alignment (AAAI 2021 · similarity 0.88)
- AXM-Net: Implicit Cross-Modal Feature Alignment for Person Re-identification (AAAI 2022 · similarity 0.88)
- Generalizable Person Re-identification via Balancing Alignment and Uniformity (NeurIPS 2024 · similarity 0.87)
- Noisy-Correspondence Learning for Text-to-Image Person Re-identification (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.