Research map / 3D vision / Human pose estimation
Temporal Feature Alignment and Mutual Information Maximization for Video-Based Human Pose Estimation
Check 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 |
|---|---|---|---|
| Apply hierarchical feature alignment and mutual-information supervision to temporal context | HRNet-W48 for video pose estimation · PoseTrack2017 validation | mAP: 82.8 -> 84.8 | Helped |
| Apply hierarchical feature alignment and mutual-information supervision to temporal context | HRNet-W48 for video pose estimation · PoseTrack2017 test | mAP: 79.2 -> 80.9 | Helped |
| Apply hierarchical feature alignment and mutual-information supervision to temporal context | HRNet-W48 for video pose estimation · PoseTrack2018 validation | mAP: 80.9 -> 82.2 | Helped |
| Apply hierarchical feature alignment and mutual-information supervision to temporal context | HRNet-W48 for video pose estimation · PoseTrack2018 test | mAP: 79.0 -> 79.6 | Helped |
| Apply hierarchical feature alignment and mutual-information supervision to temporal context | HRNet-W48 for video pose estimation · Sub-JHMDB | mAP: 94.7 -> 96.0 | Helped |
| Add global transformation for coarse feature alignment (ablation) | HRNet-W48 video pose estimation · PoseTrack2017 validation | mAP: 77.3 -> 82.9 | Helped |
| Add local calibration after global transformation (ablation) | GTM-only FAMI-Pose · PoseTrack2017 validation | mAP: 82.9 -> 84.0 | Helped |
| Add mutual-information loss to global transformation and local calibration (ablation) | GTM + LCM FAMI-Pose · PoseTrack2017 validation | mAP: 84.0 -> 84.8 | Helped |
| Increase supporting frames from N={-1} to N={-1,1} (ablation) | FAMI-Pose · PoseTrack2017 validation | mAP: 82.8 -> 83.9 | Helped |
| Add a second previous supporting frame, changing N={-1,1} to N={-2,-1,1} (ablation) | FAMI-Pose · PoseTrack2017 validation | mAP: 83.9 -> 84.5 | Helped |
| Add a second future supporting frame, changing N={-2,-1,1} to N={-2,-1,1,2} (ablation) | FAMI-Pose · PoseTrack2017 validation | mAP: 84.5 -> 84.8 | Helped |
Closest papers by meaning
- Deep Dual Consecutive Network for Human Pose Estimation (CVPR 2021 · similarity 0.92)
- Mutual Information-Based Temporal Difference Learning for Human Pose Estimation in Video (CVPR 2023 · similarity 0.91)
- Learning Temporal Pose Estimation from Sparsely-Labeled Videos (NeurIPS 2019 · similarity 0.91)
- Optimizing Human Pose Estimation Through Focused Human and Joint Regions (AAAI 2025 · similarity 0.90)
- Video-Based Human Pose Regression via Decoupled Space-Time Aggregation (CVPR 2024 · similarity 0.89)
- Can Generative Video Models Help Pose Estimation? (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.