Research map / 3D vision / Human pose estimation
Removing the Bias of Integral Pose Regression
Kerui Gu, Linlin Yang, Angela Yao
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 bias compensation to integral regression | SBL-ResNet50 IPR · COCO val | AP: 68.2 -> 71.2 | Improved |
| add bias compensation to integral regression | HRNet-W32 IPR · COCO val | AP: 72.9 -> 75.8 | Improved |
| add bias compensation to integral regression | IPR ResNet-101 · COCO val | EPE (px): 9.98 -> 9.42 | Improved |
| add bias compensation to integral regression | IPR on SBL-ResNet152 · COCO val | AP: 71.3 -> 74.4 | Improved |
| add bias compensation (joint method) | SBL-ResNet50 IPR · MPII val | PCKh@0.5: 86.2 -> 87.9 | Improved |
| add bias compensation (joint method) | HRNet-W32 IPR · MPII val | PCKh@0.5: 88.7 -> 90.6 | Improved |
| add bias compensation (joint method) | 2.5D regression · RHD test (2D) | AUC: 84.4 -> 85.8 | Improved |
| add de-bias to integral regression (ablation) | SBL-ResNet50 IPR · COCO val (ablation) | AP: 68.2 -> 71.0 | Improved |
| add detection loss Lde for first 120 epochs (ablation) | SBL-ResNet50 IPR · COCO val (ablation) | AP: 68.2 -> 70.4 | Improved |
| apply detection loss for all epochs instead of stopping (ablation) | SBL-ResNet50 IPR + Lde · COCO val (ablation) | AP: 70.4 -> 70.2 | Hurt |
| add Lde on top of de-bias (ablation) | SBL-ResNet50 IPR + de-bias · COCO val (ablation) | AP: 71.0 -> 71.2 | Improved |
| use beta=1 instead of 10 (ablation) | Full method · COCO val (ablation) | AP: 71.2 -> 70.7 | Hurt |
Closest papers by meaning
- Dive Deeper Into Integral Pose Regression (ICLR 2022 · similarity 0.94)
- Bias-Compensated Integral Regression for Human Pose Estimation (TPAMI 2023 · similarity 0.93)
- Integral Human Pose Regression (ECCV 2018 · similarity 0.90)
- Video-Based Human Pose Regression via Decoupled Space-Time Aggregation (CVPR 2024 · similarity 0.86)
- Continuous Heatmap Regression for Pose Estimation via Implicit Neural Representation (NeurIPS 2024 · similarity 0.85)
- ProbPose: A Probabilistic Approach to 2D Human Pose Estimation (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.
