Research map / 3D vision / Human pose estimation
Mutual Information-Based Temporal Difference Learning for Human Pose Estimation in Video
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 temporal-difference encoding and MI-based disentanglement | HRNet-W48 · PoseTrack2017 validation | mAP: 77.3 -> 85.7 | Helped |
| add temporal-difference encoding (ablation) | HRNet-W48 · PoseTrack2017 validation | mAP: 77.3 -> 84.5 | Helped |
| use TDE instead of optical-flow motion representations (ablation) | HRNet-W48 pose estimation · PoseTrack2017 validation | mAP: 84.0 -> 84.5 | Helped |
| add RDM with MI supervision (ablation) | HRNet-W48 + TDE · PoseTrack2017 validation | mAP: 84.5 -> 85.7 | Helped |
| use multi-stage feature differences instead of only the final-stage differences (ablation) | TDE · PoseTrack2017 validation | mAP: 84.4 -> 84.6 | Helped |
| replace simple multi-stage concatenation with progressive fusion (ablation) | TDE · PoseTrack2017 validation | mAP: 84.6 -> 84.9 | Helped |
| add spatial modulation to progressive multi-stage fusion (ablation) | TDE · PoseTrack2017 validation | mAP: 84.9 -> 85.7 | Helped |
| add learned feature factorization without the MI objective (ablation) | RDM · PoseTrack2017 validation | mAP: 84.5 -> 84.7 | Helped |
| add the MI objective to feature factorization (ablation) | RDM · PoseTrack2017 validation | mAP: 84.7 -> 85.7 | Helped |
| replace learned channel factorization with channel splitting (ablation) | RDM · PoseTrack2017 validation | mAP: 85.7 -> 83.8 | Hurt |
| replace keyframe visual features with spatiotemporal input features | TDMI · PoseTrack2017 validation | mAP: 85.7 -> 85.9 | Helped |
Closest papers by meaning
- Temporal Feature Alignment and Mutual Information Maximization for Video-Based Human Pose Estimation (CVPR 2022 · similarity 0.91)
- Deep Dual Consecutive Network for Human Pose Estimation (CVPR 2021 · similarity 0.89)
- Video-Based Human Pose Regression via Decoupled Space-Time Aggregation (CVPR 2024 · similarity 0.87)
- Optimizing Human Pose Estimation Through Focused Human and Joint Regions (AAAI 2025 · similarity 0.87)
- Motions as Queries: One-Stage Multi-Person Holistic Human Motion Capture (CVPR 2025 · similarity 0.87)
- CoMotion: Concurrent Multi-person 3D Motion (ICLR 2025 · similarity 0.86)
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.