Research map / 3D vision / Human pose estimation
15 Keypoints Is All You Need
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 |
|---|---|---|---|
| Use Pose Entailment for tracking instead of optical flow or GCN matching | KeyTrack · PoseTrack'18 Val | % IDSW: approximately half as many IDSW as the GCN and 30% less than Optical Flow | Helped |
| Add TOKS to detector-based pose estimation | HRNet · PoseTrack'18 Val | AP Total: 70.4 -> 81.6 | Helped |
| Use TOKS instead of box propagation | HRNet pose estimation · PoseTrack'18 Val | AP Total: 77.1 -> 81.6 | Helped |
| Use TOKS threshold 0.3 rather than 0.35 (ablation) | HRNet pose estimation · PoseTrack'18 Val | AP Total: 81.1 vs. 81.6 | Hurt |
| Use TOKS threshold 0.5 rather than 0.35 (ablation) | HRNet pose estimation · PoseTrack'18 Val | AP Total: 81.5 vs. 81.6 | Hurt |
| Add segment embeddings to absolute position and type embeddings (ablation) | Transformer Matching Network · PoseTrack'18 Val | Match accuracy: 72.6 -> 93.2 (ours) | Helped |
| Add type embeddings to absolute position and segment embeddings (ablation) | Transformer Matching Network · PoseTrack'18 Val | Match accuracy: 90.0 -> 93.2 (ours) | Helped |
| Replace absolute position embeddings with relative position embeddings (ablation) | Transformer Matching Network · PoseTrack'18 Val | Match accuracy: 93.2 (ours) -> 91.3 | Hurt |
| Add relative position embeddings to the full embedding scheme (ablation) | Transformer Matching Network · PoseTrack'18 Val | Match accuracy: 93.2 (ours) -> 92.0 | Hurt |
| Reduce the number of transformer blocks, hidden size, or attention heads (ablation) | Transformer Matching Network · PoseTrack'18 Val | Tracking performance | Hurt |
| Reduce matching history to a single timestep (ablation) | KeyTrack · PoseTrack'18 Val | MOTA: up to 0.3 points | Hurt |
| Use the Hungarian algorithm instead of greedy assignment (ablation) | KeyTrack · PoseTrack'18 Val | Tracking accuracy | Hurt |
Closest papers by meaning
- Detect-and-Track: Efficient Pose Estimation in Videos (CVPR 2018 · similarity 0.92)
- PoseTrack: Joint Multi-Person Pose Estimation and Tracking (CVPR 2017 · similarity 0.90)
- PoseTrack: A Benchmark for Human Pose Estimation and Tracking (CVPR 2018 · similarity 0.89)
- Optimizing Human Pose Estimation Through Focused Human and Joint Regions (AAAI 2025 · similarity 0.85)
- DriveTrack: A Benchmark for Long-Range Point Tracking in Real-World Videos (CVPR 2024 · similarity 0.85)
- ProbPose: A Probabilistic Approach to 2D Human Pose Estimation (CVPR 2025 · similarity 0.85)
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.