Research map / 3D vision / Human pose estimation
Temporal Keypoint Matching and Refinement Network for Pose Estimation and Tracking
Chunluan Zhou Zhou Ren Gang Hua
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 pose-based NMS instead of conventional NMS (ablation) | Temporal keypoint matching and refinement network · PoseTrack 2017 val | mAP: 75.2 -> 76.1 | Improved |
| add pose-based NMS instead of conventional NMS (ablation) | Temporal keypoint matching and refinement network · PoseTrack 2017 val | MOTA: 63.5 -> 64.8 | Improved |
| use OKS instead of IOU similarity for tracking (ablation) | Temporal keypoint matching and refinement network · PoseTrack 2017 val | MOTA: 64.8 -> 65.0 | No clear effect |
| use keypoint matching (TKM) instead of IOU/OKS similarity (ablation) | Temporal keypoint matching and refinement network · PoseTrack 2017 val | MOTA: 64.8 -> 67.3 | Improved |
| use keypoint-level matching (TKM) instead of whole-body matching (TBM) (ablation) | Temporal keypoint matching and refinement network · PoseTrack 2017 val | MOTA: 66.0 -> 67.3 | Improved |
| add keypoint refinement without temporal context (ablation) | Temporal keypoint matching and refinement network · PoseTrack 2017 val | mAP: 76.1 -> 76.8 | Improved |
| add temporal context to keypoint refinement (ablation) | Temporal keypoint matching and refinement network · PoseTrack 2017 val | mAP: 76.8 -> 78.2 | Improved |
| add temporal context to keypoint refinement (ablation) | Temporal keypoint matching and refinement network · PoseTrack 2017 val | MOTA: 68.1 -> 69.6 | Improved |
| use ResNeXt-101 detector instead of ResNet-101 (ablation) | Temporal keypoint matching and refinement network · PoseTrack 2017 val | MOTA: 69.6 -> 71.4 | Improved |
| use HRNet backbone instead of ResNet-152 (ablation) | Temporal keypoint matching and refinement network · PoseTrack 2017 val | mAP: 79.0 -> 79.5 | Improved |
| add pose-based NMS instead of conventional NMS (ablation) | Temporal keypoint matching and refinement network · PoseTrack 2018 val | MOTA: 63.7 -> 65.3 | Improved |
| use keypoint matching (TKM) instead of IOU similarity (ablation) | Temporal keypoint matching and refinement network · PoseTrack 2018 val | MOTA: 65.3 -> 67.0 | Improved |
Closest papers by meaning
- Detect-and-Track: Efficient Pose Estimation in Videos (CVPR 2018 · similarity 0.92)
- 15 Keypoints Is All You Need (CVPR 2020 · similarity 0.92)
- Self-supervised Keypoint Correspondences for Multi-Person Pose Estimation and Tracking in Videos (ECCV 2020 · similarity 0.91)
- Deep Dual Consecutive Network for Human Pose Estimation (CVPR 2021 · similarity 0.91)
- PoseTrack: Joint Multi-Person Pose Estimation and Tracking (CVPR 2017 · similarity 0.90)
- CoMotion: Concurrent Multi-person 3D Motion (ICLR 2025 · 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.
