Research map / 3D vision / Human pose estimation
Towards 3D Human Pose Estimation in the Wild: A Weakly-Supervised Approach
Xingyi Zhou, Qixing Huang, Xiao Sun, Xiangyang Xue, Yichen Wei
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 in-the-wild 2D-labeled data to joint training (ablation) | 3D pose network without geo loss · Human3.6M (MPJPE, S9/S11) | MPJPE (mm): 82.44 -> 65.69 | Improved |
| add geometric loss on 3D-only training (ablation) | 3D-only network · Human3.6M (MPJPE, S9/S11) | MPJPE (mm): 82.44 -> 80.98 | Improved |
| add geometric loss on 2D data (ablation) | 3D+2D network · Human3.6M (MPJPE, S9/S11) | MPJPE (mm): 65.69 -> 64.90 | Improved |
| add 2D data (with geo loss) (ablation) | 3D/w geo network · Human3.6M (MPJPE, S9/S11) | MPJPE (mm): 80.98 -> 64.90 | Improved |
| add 2D data to training (ablation) | 3D-only network · Human3.6M 2D pose | PCKh@0.5 (%): 90.01% -> 90.93% | No clear effect |
| add geometric loss (ablation) | 3D+2D network · Human3.6M 2D pose | PCKh@0.5 (%): 90.93% -> 91.62% | Improved |
| add geometric loss on 3D-only training (ablation) | 3D-only network · MPI-INF-3DHP | PCK (all): 31.5 -> 37.7 | Improved |
| add geometric loss on 2D data (ablation) | 3D+2D network · MPI-INF-3DHP | PCK (all): 65.8 -> 69.2 | Improved |
| add geometric loss on 2D data (ablation) | 3D+2D network · MPI-INF-3DHP | AUC: 32.1 -> 32.5 | Improved |
| add 2D data to training (ablation) | 3D-only network · MPI-INF-3DHP | PCK (all): 31.5 -> 65.8 | Improved |
| add geometric loss on 2D data (ablation) | 3D+2D network · MPI-INF-3DHP (Outdoor) | PCK: 67.5 -> 72.7 | Improved |
| add geometric loss on 2D data (ablation) | 3D+2D network · MPI-INF-3DHP symmetry | lower arm L1 difference (mm): 60.4mm -> 50.7mm | Improved |
Closest papers by meaning
- Weakly-Supervised 3D Human Pose Learning via Multi-View Images in the Wild (CVPR 2020 · similarity 0.94)
- In the Wild Human Pose Estimation Using Explicit 2D Features and Intermediate 3D Representations (CVPR 2019 · similarity 0.93)
- Learning 3D Human Pose from Structure and Motion (ECCV 2018 · similarity 0.92)
- 3D Human Pose Estimation in the Wild by Adversarial Learning (CVPR 2018 · similarity 0.91)
- Deductive Learning for Weakly-Supervised 3D Human Pose Estimation via Uncalibrated Cameras (AAAI 2021 · similarity 0.91)
- Self-Supervised 3D Human Pose Estimation via Part Guided Novel Image Synthesis (CVPR 2020 · similarity 0.90)
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.
