Research map / 3D vision / Human pose estimation
Rethinking the Heatmap Regression for Bottom-Up Human Pose Estimation
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 SAHR and WAHR | HrHRNet-W48 · COCO test-dev2017, multi-scale test | AP: 70.5 -> 72.0 | Helped |
| Add SAHR and WAHR | HrHRNet-W48 · CrowdPose test, multi-scale test | AP: +6.2AP | Helped |
| Add SAHR only (ablation) | HrHRNet-W32 · COCO val2017 | AP: 67.1 -> 67.8 | Helped |
| Add WAHR only (ablation) | HrHRNet-W32 · COCO val2017 | AP: 67.1 -> 68.4 | Helped |
| Add WAHR to SAHR (ablation) | HrHRNet-W32 + SAHR · COCO val2017 | AP: 67.8 -> 68.9 | Helped |
| Replace learned SAHR scale maps with bounding-box-derived scale factors (ablation) | HrHRNet-W32 · COCO val2017 | AP: 67.1 -> 63.9 | Hurt |
| Set SAHR regularizer weight λ to 0.5 instead of 1.0 (ablation) | HrHRNet-W32 + SAHR · COCO val2017 | AP: 67.8 -> 67.8 | No clear effect |
| Reduce SAHR regularizer weight λ to 0.1 (ablation) | HrHRNet-W32 + SAHR · COCO val2017 | AP: 67.8 -> 67.6 | Hurt |
| Reduce WAHR γ from 1.0 to 0.01 (ablation) | HrHRNet-W32 + WAHR · COCO val2017 | AP: 67.8 -> 68.4 | Helped |
| Reduce WAHR γ from 0.01 to 0.001 (ablation) | HrHRNet-W32 + WAHR · COCO val2017 | AP: 68.4 -> 68.4 | No clear effect |
| Increase dilation rates from 1111 to 2222 (ablation) | HrHRNet-W32 · COCO val2017 | AP: 66.6 -> 66.7 | No clear effect |
| Increase Gaussian standard deviation from σ=2 to σ=3 (ablation) | HrHRNet-W32 · COCO val2017, medium and large persons | AP: APᴹ: 61.3 -> 58.3; APᴸ: 75.0 -> 75.4 | Mixed |
Closest papers by meaning
- Continuous Heatmap Regression for Pose Estimation via Implicit Neural Representation (NeurIPS 2024 · similarity 0.86)
- ProbPose: A Probabilistic Approach to 2D Human Pose Estimation (CVPR 2025 · similarity 0.85)
- PoseGen: Learning to Generate 3D Human Pose Dataset with NeRF (AAAI 2024 · similarity 0.84)
- SHaRPose: Sparse High-Resolution Representation for Human Pose Estimation (AAAI 2024 · similarity 0.84)
- Reconstructing People, Places, and Cameras (CVPR 2025 · similarity 0.84)
- MotionMap: Representing Multimodality in Human Pose Forecasting (CVPR 2025 · similarity 0.84)
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.