Research map / Label-efficient and robust learning / Recognition and long-tail
Guided Saliency Feature Learning for Person Re-identification in Crowded Scenes
Lingxiao He, Wu Liu
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 mask layer to FCN baseline (ablation) | FCN + average pooling · Crowd REID (trained on Market1501) | rank-1 accuracy (%): 16.7 -> 32.2 | Improved |
| add pose layer to FCN baseline (ablation) | FCN + average pooling · Crowd REID (trained on Market1501) | rank-1 accuracy (%): 16.7 -> 30.0 | Improved |
| combine pose layer and mask layer (ablation) | FCN + average pooling · Crowd REID (trained on Market1501) | rank-1 accuracy (%): 16.7 -> 32.4 | Improved |
| add pose and mask layers (ablation) | FCN + average pooling · Occluded REID (trained on Market1501) | rank-1 accuracy (%): 60.2 -> 74.5 | Improved |
| add mask layer (ablation) | FCN + average pooling · Market1501 (trained on Market1501) | rank-1 accuracy (%): 94.8 -> 94.5 | No clear effect |
| add pose and mask layers (ablation) | FCN + average pooling · Market1501 (trained on Market1501) | rank-1 accuracy (%): 94.8 -> 95.3 | Improved |
| add pose and mask layers plus guided matching | FCN baseline · Market1501 | mAP (%): 83.4 -> 84.7 | Improved |
| add GASM to baseline for cross-domain transfer | ResNet-50 baseline · Cross-domain Market1501 to MSMT17 | rank-1 accuracy (%): 22.2 -> 27.8 | Improved |
| add GASM to baseline for cross-domain transfer | ResNet-50 baseline · Cross-domain DukeMTMC to Market1501 | rank-1 accuracy (%): 61.6 -> 67.2 | Improved |
| add GASM to baseline for cross-domain transfer | ResNet-50 baseline · Cross-domain Market1501 to DukeMTMC | rank-1 accuracy (%): 54.3 -> 56.4 | Improved |
| add GASM, trained on MSMT17 | ResNet-50 baseline · Crowd REID (trained on MSMT17) | rank-1 accuracy (%): 44.0 -> 51.5 | Improved |
| compare GASM with FPR | FPR · DukeMTMC-reID | rank-1 accuracy (%): 88.6 -> 88.3 | Hurt |
Closest papers by meaning
- Pose-Guided Feature Alignment for Occluded Person Re-Identification (ICCV 2019 · similarity 0.90)
- Person Re-Identification by Saliency Learning (TPAMI 2016 · similarity 0.90)
- Occlude Them All: Occlusion-Aware Attention Network for Occluded Person Re-ID (ICCV 2021 · similarity 0.89)
- Occluded Person Re-identification via Saliency-Guided Patch Transfer (AAAI 2024 · similarity 0.89)
- Dynamic Feature Pruning and Consolidation for Occluded Person Re-identification (AAAI 2024 · similarity 0.88)
- From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · similarity 0.88)
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.
