Research map / Label-efficient and robust learning / Recognition and long-tail
Second-Order Non-Local Attention Networks for Person Re-Identification
Bryan (Ning) Xia, Yuan Gong, Yizhe Zhang, Christian Poellabauer
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 SONA after stage 2 (ablation) | baseline with branches · CUHK03 labeled (new protocol) | mAP: 77.05 -> 78.48 | Improved |
| add SONA after stage 2 (ablation) | baseline with branches · Market1501 | mAP: 87.50 -> 88.50 | Improved |
| add SONA after stage 2 (ablation) | baseline with branches · DukeMTMC-reID | mAP: 76.2 -> 78.18 | Improved |
| add DropBlock+ to baseline (ablation) | baseline · CUHK03 labeled (new protocol) | mAP: 77.05 -> 77.45 | No clear effect |
| add DropBlock+ to baseline (ablation) | baseline · Market1501 | mAP: 87.50 -> 87.68 | No clear effect |
| add original DropBlock to baseline (ablation) | baseline · CUHK03 labeled (new protocol) | mAP: 77.05 -> 77.02 | No clear effect |
| add BFE batch erasing to baseline (ablation) | baseline · CUHK03 labeled (new protocol) | mAP: 77.05 -> 77.20 | No clear effect |
| place SONA after stage 4 instead of early stages (ablation) | SONA-Net · CUHK03 labeled (new protocol) | mAP: 75.8 -> 79.23 | Hurt |
| apply SONA after stage 2, stage 3 or both (ablation) | SONA-Net · Market1501 | mAP: 88.67 vs 88.63 vs 88.83 | No clear effect |
| add SONA module overhead at inference (ablation) | model with SONA2 · Market1501 | inference time (ms): 7.89 -> 8.44 | No clear effect |
| add DropBlock+ to baseline: training loss (ablation) | baseline · CUHK03 labeled (new protocol) | average training loss: +0.45% | Mixed |
Closest papers by meaning
- Relation-Aware Global Attention for Person Re-Identification (CVPR 2020 · similarity 0.91)
- Harmonious Attention Network for Person Re-Identification (CVPR 2018 · similarity 0.89)
- Scalable Person Re-Identification by Harmonious Attention (IJCV 2019 · similarity 0.89)
- Discriminative Feature Learning With Consistent Attention Regularization for Person Re-Identification (ICCV 2019 · similarity 0.88)
- Mixed High-Order Attention Network for Person Re-Identification (ICCV 2019 · similarity 0.88)
- Beyond Human Parts: Dual Part-Aligned Representations for Person Re-Identification (ICCV 2019 · 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.
