Research map / 3D vision / Human pose estimation
Human Pose As Compositional Tokens
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 |
|---|---|---|---|
| Use compositional pose tokens instead of independent joint codebooks (ablation) | PCT pose estimation · COCO val2017 | AP^P: 16.2 -> 65.5 | Helped |
| Add masked joint modeling (ablation) | PCT · COCO val2017 | AP^P: 65.5 -> 72.7 | Helped |
| Add masked joint modeling (ablation) | PCT · SyncOCC test | AP^OC: 88.3 -> 91.2 | Helped |
| Add image guidance (ablation) | PCT · COCO val2017 | AP^P: 72.7 -> 75.1 | Helped |
| Add image guidance (ablation) | PCT · SyncOCC test | AP^OC: 91.2 -> 92.8 | Helped |
| Add auxiliary pose reconstruction loss (ablation) | PCT · COCO val2017 | AP^P: 75.1 -> 77.4 | Helped |
| Add auxiliary pose reconstruction loss (ablation) | PCT · SyncOCC test | AP^OC: 92.8 -> 93.1 | Helped |
| Use PCT rather than coordinate representation | Swin-Base pose estimation · OCHuman val occluded joints | AP^OC: 41.5 -> 45.6 | Helped |
| Use PCT rather than heatmaps | Swin-Base pose estimation · OCHuman val occluded joints | AP^OC: 40.1 -> 45.6 | Helped |
| Use PCT rather than discrete bins | Swin-Base pose estimation · OCHuman val occluded joints | AP^OC: 40.5 -> 45.6 | Helped |
| Increase the number of tokens from 4 to 16 (ablation) | PCT · COCO val2017 | AP | Helped |
| Increase token count beyond 16 (ablation) | PCT · COCO val2017 | AP | No clear effect |
Closest papers by meaning
- Kinematic-Structure-Preserved Representation for Unsupervised 3D Human Pose Estimation (AAAI 2020 · similarity 0.87)
- FlexPose: Pose Distribution Adaptation with Limited Guidance (AAAI 2025 · similarity 0.86)
- Capturing the Motion of Every Joint: 3D Human Pose and Shape Estimation with Independent Tokens (ICLR 2023 · similarity 0.86)
- TokenHMR: Advancing Human Mesh Recovery with a Tokenized Pose Representation (CVPR 2024 · similarity 0.86)
- PoseBH: Prototypical Multi-Dataset Training Beyond Human Pose Estimation (CVPR 2025 · similarity 0.85)
- Video-Based Human Pose Regression via Decoupled Space-Time Aggregation (CVPR 2024 · 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.