Research map / 3D vision / Human pose estimation
H3WB: Human3.6M 3D WholeBody Dataset and Benchmark
Yue Zhu, Nermin Samet, David Picard
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 diffusion-based 2D refinement after the completion network (ablation) | H3WB annotation pipeline · manual cross-check on 600 images | 3D error all (mm): 29.31 -> 16.98 | Improved |
| add completion network to fill missing keypoints (ablation) | H3WB annotation pipeline · manual cross-check on 600 images | 3D error all (mm): 14.87 -> 29.31 | Hurt |
| train on H3WB + T3WB instead of H3WB only | SimpleBaseline · H3WB 2D->3D | MPJPE all (mm): 125.4 -> 115.3 | Improved |
| train on H3WB + T3WB instead of H3WB only | Jointformer · H3WB 2D->3D | MPJPE all (mm): 88.3 -> 81.5 | Improved |
| train on H3WB + T3WB instead of H3WB only | Jointformer · H3WB I2D->3D | MPJPE all (mm): 109.2 -> 84.2 | Improved |
| train on H3WB + T3WB instead of H3WB only | SimpleBaseline · H3WB I2D->3D | MPJPE all (mm): 268.8 -> 260.5 | No clear effect |
| enlarge SimpleBaseline from 6 to 8 layers (Large SimpleBaseline) (ablation) | SimpleBaseline · H3WB 2D->3D | MPJPE all (mm): 125.4 -> 112.3 | Improved |
| remove batch normalization / enlarge SimpleBaseline for incomplete input (ablation) | SimpleBaseline · H3WB I2D->3D | MPJPE all (mm): 268.8 -> 131.4 | Improved |
| add 3D supervision with fixed camera views to CanonPose (ablation) | CanonPose · H3WB 2D->3D | MPJPE all (mm): 186.7 -> 117.7 | Improved |
| use discriminative lifting rather than parametric model fitting (SMPLify-X) | 2D->3D lifting · H3WB 2D->3D | MPJPE all (mm): 188.9 -> 88.3 (Jointformer) | Improved |
| use a single-stage ResNet50 instead of two-stage SHN + SimpleBaseline | RGB->3D · H3WB RGB->3D | MPJPE all (mm): 182.5 -> 166.7 | Improved |
| use part-disjoint DOPE model rather than jointly trained whole-body models | RGB->3D · H3WB RGB->3D | MPJPE all (mm): 132.6 -> 191.3 | Hurt |
Closest papers by meaning
- Whole-Body Human Pose Estimation in the Wild (ECCV 2020 · similarity 0.91)
- DOPE: Distillation Of Part Experts for whole-body 3D pose estimation in the wild (ECCV 2020 · similarity 0.89)
- SAM 3D Body: Robust Full-Body Human Mesh Recovery (CVPR 2026 · similarity 0.89)
- Motion-X: A Large-scale 3D Expressive Whole-body Human Motion Dataset (NeurIPS 2023 · similarity 0.89)
- Lifting 2D Human Pose to 3D with Domain Adapted 3D Body Concept (IJCV 2023 · similarity 0.89)
- Forecasting of 3D Whole-body Human Poses with Grasping Objects (CVPR 2024 · 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.
