Research map / 3D vision / Human pose estimation
3D Human Pose Estimation From a Single Image via Distance Matrix Regression
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| represent 2D and 3D poses as pairwise distance matrices and regress between them | 3D human pose estimation · Humaneva-I Walking | average joint error (mm): 36.60 -> 19.20 | Helped |
| represent poses as pairwise distance matrices and regress them | 3D human pose estimation · Human3.6M Protocol #1 | average joint error (mm): 112.91 -> 87.30 | Helped |
| represent poses as pairwise distance matrices and regress them | 3D human pose estimation · Human3.6M Protocol #2 | average joint error (mm): 82.03 vs 83.52 | Mixed |
| train on GT+CPM 2D detections instead of ground-truth 2D alone (ablation) | FConv on Humaneva-I · Humaneva-I Walking | average joint error (mm): 31.68 -> 19.20 | Helped |
| train on CPM detections instead of GT+CPM (ablation) | FConn on Humaneva-I · Humaneva-I Walking | average joint error (mm): 19.90 -> 20.97 | Hurt |
| use FConn instead of FConv (ablation) | Humaneva-I regressor trained on CPM · Humaneva-I Boxing | average joint error (mm): 46.57 -> 49.86 | Hurt |
| add Gaussian noise to 2D ground-truth joint positions (ablation) | FConv on Human3.6M · Human3.6M Protocol #3 noise test | average joint error (mm): 62.17 -> 67.11 -> 79.12 -> 96.08 -> 115.55 | Hurt |
| train with synthetic joint occlusions to predict missing joints (ablation) | FConv · Humaneva-I, two random occluded joints | average joint error (mm): 32.68 vs 54.93 | Helped |
| train with synthetic joint occlusions to predict missing joints (ablation) | FConv · Human3.6M Protocol #3, two random occluded joints | average joint error (mm): 106.79 | Mixed |
Closest papers by meaning
- 3D Human Pose Estimation = 2D Pose Estimation + Matching (CVPR 2017 · similarity 0.92)
- Coarse-To-Fine Volumetric Prediction for Single-Image 3D Human Pose (CVPR 2017 · similarity 0.91)
- 3D Human Pose Estimation in the Wild by Adversarial Learning (CVPR 2018 · similarity 0.91)
- Sparseness Meets Deepness: 3D Human Pose Estimation From Monocular Video (CVPR 2016 · similarity 0.91)
- Kinematic-Structure-Preserved Representation for Unsupervised 3D Human Pose Estimation (AAAI 2020 · similarity 0.91)
- Neural Localizer Fields for Continuous 3D Human Pose and Shape Estimation (NeurIPS 2024 · 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.