Research map / 3D vision / Human pose estimation
3D Human Pose Estimation via Explicit Compositional Depth Maps
Haiping Wu, Bin Xiao
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 Limb Depth Maps instead of Joint Depth Maps as depth target (ablation) | ResNet-50 pipeline · Human3.6M | MPJPE (Z): 37.9 -> 36.3 | Improved |
| use Limb Depth Maps instead of Joint Depth Maps as depth target (ablation) | ResNet-50 pipeline · Human3.6M | P-MPJPE: 41.0 -> 39.3 | Improved |
| add 2D heatmap supervision to direct regression (ablation) | ResNet-50 pipeline · Human3.6M | MPJPE (X,Y): 30.1 -> 27.7 | Improved |
| add 2D heatmap supervision to direct regression (ablation) | ResNet-50 pipeline · Human3.6M | MPJPE (Z): 37.5 -> 37.9 | Hurt |
| add limb depth maps on top of 2D heatmaps (explicit depth for lifting) (ablation) | ResNet-50 pipeline · Human3.6M | MPJPE: 51.7 -> 48.8 | Improved |
| add 2D heatmaps and limb depth maps to direct regression (ablation) | ResNet-50 pipeline · Human3.6M | MPJPE: 52.9 -> 48.8 | Improved |
| replace single-frame state of the art with HRNet-W32 version of method (ablation) | Ours · Human3.6M | MPJPE: 47.3 -> 43.2 | Improved |
| compare to 243-frame temporal method using a single frame | Ours (HRNet-W32) · Human3.6M | MPJPE: 46.8 -> 43.2 | Mixed |
Closest papers by meaning
- Sparseness Meets Deepness: 3D Human Pose Estimation From Monocular Video (CVPR 2016 · similarity 0.89)
- Consensus-Based Optimization for 3D Human Pose Estimation in Camera Coordinates (IJCV 2022 · similarity 0.89)
- Learning to Fuse 2D and 3D Image Cues for Monocular Body Pose Estimation (ICCV 2017 · similarity 0.89)
- Monocular 3D Human Pose Estimation by Predicting Depth on Joints (ICCV 2017 · similarity 0.88)
- A Simple yet Effective Baseline for 3D Human Pose Estimation (ICCV 2017 · similarity 0.88)
- PandaPose: 3D Human Pose Lifting from a Single Image via Propagating 2D Pose Prior to 3D Anchor Space (NeurIPS 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.
