Research map / Label-efficient and robust learning / Recognition and long-tail
Beyond Human Parts: Dual Part-Aligned Representations for Person Re-Identification
Jianyuan Guo, Yuhui Yuan, Lang Huang, Chao Zhang, Jin-Ge Yao, Kai Han
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 |
|---|---|---|---|
| add 5 dual part-aligned blocks (DPB) | ResNet-50 baseline · Market-1501 (softmax only) | mAP: 71.48 -> 83.40 | Improved |
| add latent part branch to human part branch (ablation) | 5 x DPB (HP-5 only) · Market-1501 (softmax only) | mAP: 79.28 -> 83.40 | Improved |
| add human part branch to latent part branch (ablation) | 5 x DPB (latent only) · Market-1501 (softmax only) | mAP: 80.49 -> 83.40 | Improved |
| increase DPB blocks from 1 to 5 (ablation) | DPB (HP-5 + Latent) · Market-1501 (softmax only) | mAP: 80.98 -> 83.40 | Improved |
| use K=1 human part (no parsing information) (ablation) | DPB (HP-1, Res-1) · Market-1501 (softmax only) | Rank-1: 88.36 -> 88.98 | No clear effect |
| increase human parts from K=2 to K=5 (ablation) | DPB human part branch (Res-2) · Market-1501 (softmax only) | mAP: 75.87 -> 78.72 | Improved |
| restrict latent branch attention to non-human regions rather than human regions (ablation) | DPB latent branch (Res-2) · Market-1501 (softmax only) | mAP: 77.22 -> 78.25 | Improved |
| insert DPB after Res-2 instead of Res-4 (ablation) | DPB (HP-5 + Latent) · Market-1501 (softmax only) | mAP: 78.36 -> 80.98 | Improved |
| add triplet loss to P2-Net | P2-Net · Market-1501 | mAP: 83.4 -> 85.6 | Improved |
| add triplet loss to P2-Net | P2-Net · DukeMTMC-reID | mAP: 70.8 -> 73.1 | Improved |
| add triplet loss to P2-Net | P2-Net · CUHK03 (detected) | mAP: 64.2 -> 68.9 | Improved |
| add latent branch to human part branch with 5 blocks (ablation) | 5 x DPB (HP-5) · DukeMTMC-reID | mAP: 68.64 -> 70.84 | Improved |
Closest papers by meaning
- Human Semantic Parsing for Person Re-Identification (CVPR 2018 · similarity 0.91)
- Deeply-Learned Part-Aligned Representations for Person Re-Identification (ICCV 2017 · similarity 0.91)
- Identity-Guided Human Semantic Parsing for Person Re-Identification (ECCV 2020 · similarity 0.91)
- Multi-Task Learning With Coarse Priors for Robust Part-Aware Person Re-Identification (TPAMI 2020 · similarity 0.89)
- Part-Aligned Bilinear Representations for Person Re-Identification (ECCV 2018 · similarity 0.89)
- Dynamic Feature Pruning and Consolidation for Occluded Person Re-identification (AAAI 2024 · similarity 0.87)
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.
