Research map / 3D vision / Human pose estimation
Bottom-Up Human Pose Estimation via Disentangled Keypoint Regression
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 |
|---|---|---|---|
| Add adaptive activation to the regression baseline (ablation) | DEKR regression head · COCO val2017 | AP: 61.9 -> 65.4; +3.5 | Helped |
| Add separate regression without adaptive activation (ablation) | DEKR regression head · COCO val2017 | AP: 61.9 -> 63.6 | Helped |
| Add separate regression to adaptive activation (ablation) | DEKR regression head · COCO val2017 | AP: 65.4 -> 68.0; +2.6 | Helped |
| Add adaptive activation and separate regression (ablation) | DEKR regression head · COCO val2017 | Jitter error: 16.4 -> 11.8 | Helped |
| Add adaptive activation and separate regression (ablation) | DEKR regression head · COCO val2017 | Miss error: 7.6 -> 6.1 | Helped |
| Group multiple keypoints into shared regression branches instead of one branch per keypoint (ablation) | DEKR separate regression · COCO val2017, HRNet-W32 | AP: decreases by 0.4 | Hurt |
| Match regressed keypoints to closest heatmap-detected keypoints (ablation) | DEKR · COCO validation and test, single-scale | AP: 68.0_{-0.0}; 71.0_{-0.0}; 67.3_{-0.0}; 70.1_{\textcolor{red}{\uparrow 0.1}} | No clear effect |
| Match regressed keypoints to closest heatmap-detected keypoints (ablation) | DEKR · COCO validation and test, multi-scale | AP: 71.0_{ \textcolor{red}{\uparrow 0.3}}; 72.8_{\textcolor{red}{\uparrow 0.5}}; 70.2_{\textcolor{red}{\uparrow 0.4}}; 71.4_ | Helped |
| Match regressed keypoints to closest heatmap-detected keypoints (ablation) | DEKR · CrowdPose test, single-scale | AP: 65.5_{\textcolor{blue}{\downarrow 0.2}}; 67.0_{\textcolor{blue}{\downarrow 0.3}} | Hurt |
| Match regressed keypoints to closest heatmap-detected keypoints (ablation) | DEKR · CrowdPose test, multi-scale | AP: 67.5_{\textcolor{red}{\uparrow 0.5}}; 68.3_{\textcolor{red}{\uparrow 0.3}} | Helped |
| Use multi-scale rather than single-scale testing | DEKR (HRNet-W32) · COCO val2017 | AP: +2.7 | Helped |
| Use multi-scale rather than single-scale testing | DEKR (HRNet-W48) · COCO val2017 | AP: +1.3 | Helped |
Closest papers by meaning
- QueryPose: Sparse Multi-Person Pose Regression via Spatial-Aware Part-Level Query (NeurIPS 2022 · similarity 0.89)
- ProbPose: A Probabilistic Approach to 2D Human Pose Estimation (CVPR 2025 · similarity 0.88)
- Meta-Point Learning and Refining for Category-Agnostic Pose Estimation (CVPR 2024 · similarity 0.86)
- SHaRPose: Sparse High-Resolution Representation for Human Pose Estimation (AAAI 2024 · similarity 0.86)
- Optimizing Human Pose Estimation Through Focused Human and Joint Regions (AAAI 2025 · similarity 0.86)
- ESCAPE: Encoding Super-keypoints for Category-Agnostic Pose Estimation (CVPR 2024 · 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.