Research map / Video / Video segmentation and tracking
Diversity Regularized Spatiotemporal Attention for Video-Based Person Re-Identification
Read the paperCheck 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 diverse spatial attention models (ablation) | ResNet-50 video re-identification baseline · PRID2011 | rank-1 accuracy: 82.7 -> 84.2 | Helped |
| Replace Q' with Hellinger-distance regularization Q (ablation) | SpaAtn · iLIDS-VID | rank-1 accuracy: 64.5 -> 68.6 | Helped |
| Use Hellinger-distance regularization Q instead of no diversity regularizer (ablation) | SpaAtn · MARS | rank-1 accuracy: 74.5 -> 77.0 | Helped |
| Use max temporal pooling instead of average temporal pooling (ablation) | SpaAtn+Q · iLIDS-VID | rank-1 accuracy: 68.6 -> 68.2 | Hurt |
| Add per-region temporal attention (ablation) | SpaAtn+Q · PRID2011 | rank-1 accuracy: 86.7 -> 88.4 | Helped |
| Fine-tune the entire network independently on each dataset (ablation) | SpaAtn+Q+TemAtn · iLIDS-VID | rank-1 accuracy: 69.7 -> 80.2 | Helped |
| Vary the number of spatial attention models from K=1 to K=2 (ablation) | Proposed framework · PRID2011 | rank-1 accuracy: 86.2 -> 83.4 | Hurt |
| Increase the number of spatial attention models from K=2 to K=6 (ablation) | Proposed framework · iLIDS-VID | rank-1 accuracy: 64.6 -> 69.7 | Helped |
| Use K=6 rather than K=4 spatial attention models (ablation) | Proposed framework · MARS | rank-1 accuracy: 77.2 -> 77.1 | Hurt |
| Use spatiotemporal attention with independent dataset fine-tuning | Person re-identification methods · MARS | mAP: 50.7 -> 65.8 | Helped |
Closest papers by meaning
- STA: Spatial-Temporal Attention for Large-Scale Video-Based Person Re-Identification (AAAI 2019 · similarity 0.91)
- Spatial-Temporal Graph Convolutional Network for Video-Based Person Re-Identification (CVPR 2020 · similarity 0.90)
- Relation-Aware Global Attention for Person Re-Identification (CVPR 2020 · similarity 0.90)
- BiCnet-TKS: Learning Efficient Spatial-Temporal Representation for Video Person Re-Identification (CVPR 2021 · similarity 0.89)
- Relation-Guided Spatial Attention and Temporal Refinement for Video-Based Person Re-Identification (AAAI 2020 · similarity 0.89)
- Exploring the Camera Bias of Person Re-identification (ICLR 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.