Research map / 3D vision / Human pose estimation
DOPE: Distillation Of Part Experts for whole-body 3D pose estimation in the wild
Philippe Weinzaepfel, Romain Brégier, Hadrien Combaluzier, Vincent Leroy, Grégory Rogez
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 expert pseudo ground truth for missing parts (DOPE without distillation loss vs gt-only) (ablation) | whole-body network · RenderedHand test | AUC: 81.1 -> 83.5 | Improved |
| add expert pseudo ground truth for missing parts (ablation) | whole-body network · Menpo 3D tracking test | AUC: 61.7 -> 75.2 | Improved |
| add expert pseudo ground truth for missing parts (ablation) | whole-body network · MPII (1000 held-out train images) | PCKh@0.5: 88.3 -> 88.3 | No clear effect |
| add expert pseudo ground truth for missing parts (ablation) | whole-body network · MuPoTs-3D | PCK3D: 66.6 -> 66.4 | No clear effect |
| add distillation loss (ablation) | whole-body network · RenderedHand test | AUC: 83.5 -> 84.9 | Improved |
| add distillation loss (ablation) | whole-body network · MPII (1000 held-out train images) | PCKh@0.5: 88.3 -> 88.8 | Improved |
| add distillation loss (ablation) | whole-body network · MuPoTs-3D | PCK3D: 66.4 -> 67.2 | Improved |
| add distillation loss (ablation) | whole-body network · Menpo 3D tracking test | AUC: 75.2 -> 75.0 | No clear effect |
| reduce image size to 400px, 50 proposals, half precision | DOPE · Inference speed (NVIDIA T4) | runtime per image (ms): 100 -> 28 | Mixed |
| train on union of partial datasets (unannotated parts treated as negatives) (ablation) | whole-body network · hand and face detection | performance drop: ~10% drop for hands and faces, ~2% for bodies | Hurt |
Closest papers by meaning
- Whole-Body Human Pose Estimation in the Wild (ECCV 2020 · similarity 0.90)
- Single-Network Whole-Body Pose Estimation (ICCV 2019 · similarity 0.90)
- ZoomNAS: Searching for Whole-body Human Pose Estimation in the Wild (TPAMI 2022 · similarity 0.90)
- H3WB: Human3.6M 3D WholeBody Dataset and Benchmark (ICCV 2023 · similarity 0.89)
- Monocular Total Capture: Posing Face, Body, and Hands in the Wild (CVPR 2019 · similarity 0.88)
- Multi-HMR: Multi-Person Whole-Body Human Mesh Recovery in a Single Shot (ECCV 2024 · similarity 0.88)
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.
