Research map / 3D vision / Human pose estimation
Pose-Oriented Transformer with Uncertainty-Guided Refinement for 2D-to-3D Human Pose Estimation
Han Li, Bowen Shi, Wenrui Dai, Hongwei Zheng, Botao Wang, Yu Sun, Min Guo, Chenglin Li, Junni Zou, Hongkai Xiong
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 |
|---|---|---|---|
| Use the full POT and uncertainty-guided refinement method | 3D human pose estimation · Human3.6M test (CPN 2D input) | MPJPE (mm): 51.9 -> 50.5 | Improved |
| Use the full POT and uncertainty-guided refinement method | 3D human pose estimation · Human3.6M test (GT 2D input) | MPJPE (mm): 35.2 -> 33.8 | Improved |
| Use the full method trained on Human3.6M | 3D human pose estimation · MPI-INF-3DHP test | PCK: 80.1 -> 84.1 | Improved |
| Use the full method trained on Human3.6M | 3D human pose estimation · MPI-INF-3DHP test | AUC: 45.8 -> 53.7 | Improved |
| Add distance-related group position embeddings (ablation) | POT · Human3.6M ablation (GT 2D input) | MPJPE (mm): 37.57 -> 36.69 | Improved |
| Replace standard self-attention with PO-SA (ablation) | POT · Human3.6M ablation (GT 2D input) | MPJPE (mm): 36.69 -> 35.59 | Improved |
| Add UGRN refinement (ablation) | POT · Human3.6M ablation (GT 2D input) | MPJPE (mm): 35.59 -> 34.72 | Improved |
| Add uncertainty-guided sampling (ablation) | POT + UGRN · Human3.6M ablation (GT 2D input) | MPJPE (mm): 34.72 -> 33.82 | Improved |
| Use UG-SA instead of standard MH-SA (ablation) | UGRN · Human3.6M ablation (GT 2D input) | MPJPE (mm): 35.22 -> 34.72 | Improved |
| Use UG-SA instead of PO-SA (ablation) | UGRN · Human3.6M ablation (GT 2D input) | MPJPE (mm): 35.07 -> 34.72 | Improved |
| Increase POT and UGRN depth from 4 + 1 to 8 + 2 layers (ablation) | POT + UGRN · Human3.6M ablation (GT 2D input) | MPJPE (mm): 37.08 -> 35.20 | Improved |
| Increase POT and UGRN depth from 8 + 2 to 12 + 3 layers (ablation) | POT + UGRN · Human3.6M ablation (GT 2D input) | MPJPE (mm): 35.20 -> 33.82 | Improved |
Closest papers by meaning
- PoseFormerV2: Exploring Frequency Domain for Efficient and Robust 3D Human Pose Estimation (CVPR 2023 · similarity 0.89)
- 3D Human Pose Estimation From a Single Image via Distance Matrix Regression (CVPR 2017 · similarity 0.88)
- HopFIR: Hop-wise GraphFormer with Intragroup Joint Refinement for 3D Human Pose Estimation (ICCV 2023 · similarity 0.88)
- Self-Supervised 3D Human Mesh Recovery from a Single Image with Uncertainty-Aware Learning (AAAI 2024 · similarity 0.87)
- Deep Semantic Graph Transformer for Multi-View 3D Human Pose Estimation (AAAI 2024 · similarity 0.87)
- Reconstructing Humans with a Biomechanically Accurate Skeleton (CVPR 2025 · similarity 0.87)
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.
