Research map / Architectures and efficiency / Attention and transformers
Self-Critical Attention Learning for Person Re-Identification
Guangyi Chen, Chunze Lin, Liangliang Ren, Jiwen Lu, Jie Zhou
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 self-critical module to spatial attention (ablation) | spatial attention ResNet-50 · Market-1501 | mAP: 87.6 -> 88.9 | Improved |
| add self-critical module to channel attention (ablation) | channel attention ResNet-50 · Market-1501 | mAP: 88.1 -> 89.3 | Improved |
| add triplet loss (ablation) | ResNet-50 CE baseline · Market-1501 | mAP: +1.9% | Improved |
| add label smoothing regularization (ablation) | ResNet-50 CE+triplet · Market-1501 | mAP: +1.4% | Improved |
| add horizontal flip feature averaging at inference (ablation) | ResNet-50 CE · Market-1501 | mAP: 82.1 -> 82.2 | Improved |
| add spatial attention (ablation) | ResNet-50 CE+flip+triplet+LSR · Market-1501 | mAP: 85.5 -> 87.6 | Improved |
| add channel attention (ablation) | ResNet-50 CE+flip+triplet+LSR · Market-1501 | mAP: 85.5 -> 88.1 | Improved |
| use SCAL (spatial) instead of baseline in cross-domain transfer (ablation) | ResNet-50 baseline · Market-1501 to DukeMTMC-ReID | mAP: 13.1 -> 17 | Improved |
| use SCAL (channel) instead of baseline in cross-domain transfer (ablation) | ResNet-50 baseline · DukeMTMC-ReID to Market-1501 | mAP: 18.8 -> 23.8 | Improved |
| replace SPGAN feature extractor with SCAL (channel) (ablation) | SPGAN · Market-1501 to DukeMTMC-ReID | mAP: 22.3 -> 28.4 | Improved |
| choose spatial vs channel attention | SCAL · Market-1501, CUHK03 vs DukeMTMC-ReID | mAP: 89.3 vs 88.9 (Market-1501); 79.6 vs 79.1 (DukeMTMC-ReID) | Mixed |
Closest papers by meaning
- Mixed High-Order Attention Network for Person Re-Identification (ICCV 2019 · similarity 0.89)
- Mancs: A Multi-task Attentional Network with Curriculum Sampling for Person Re-identification (ECCV 2018 · similarity 0.88)
- Relation-Aware Global Attention for Person Re-Identification (CVPR 2020 · similarity 0.87)
- Scalable Person Re-Identification by Harmonious Attention (IJCV 2019 · similarity 0.87)
- Re-Identification With Consistent Attentive Siamese Networks (CVPR 2019 · similarity 0.87)
- Harmonious Attention Network for Person Re-Identification (CVPR 2018 · 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.
