Research map / 3D vision / Human pose estimation
Learning From Synthetic Humans
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 |
|---|---|---|---|
| Train on synthetic images instead of the limited real training set | Stacked-hourglass part segmentation · Freiburg Sitting People | mean part IOU: 28.77 -> 40.10 | Helped |
| Fine-tune the synthetic-trained network on real images (ablation) | Stacked-hourglass part segmentation · Freiburg Sitting People | mean part IOU: 40.10 -> 59.58 | Helped |
| Add a full-resolution upsampling layer (ablation) | Synth+Real part segmentation · Freiburg Sitting People | mean part IOU: 59.58 -> 68.84 | Helped |
| Fine-tune the synthetic-trained network on real images (ablation) | Stacked-hourglass part segmentation · Human3.6M | part IOU (fg+bg): 46.35 -> 57.07 | Helped |
| Train on synthetic images rather than real images only | Stacked-hourglass part segmentation · Human3.6M | part IOU (fg): 46.32 -> 42.91 | Hurt |
| Fine-tune the synthetic-trained network on real images (ablation) | Stacked-hourglass depth estimation · Human3.6M | depth RMSE: 111.6 -> 90.0 | Helped |
| Use Human3.6M rather than CMU MoCap to generate synthetic pretraining data (ablation) | Synthetic-pretrained segmentation and depth networks · Human3.6M | segmentation IOU: 42.82% -> 48.11% | Helped |
| Use Human3.6M rather than CMU MoCap to generate synthetic pretraining data (ablation) | Synthetic-pretrained depth network · Human3.6M | depth RMSE: 2.57 -> 2.44 | Helped |
| Increase the amount of synthetic training data (ablation) | SURREAL-trained segmentation and depth networks · SURREAL and Human3.6M test sets | performance | Helped |
| Increase clothing variation in synthetic training clips (ablation) | SURREAL-trained segmentation and depth networks · SURREAL and Human3.6M test sets | performance | Helped |
Closest papers by meaning
- Sim2real transfer learning for 3D human pose estimation: motion to the rescue (NeurIPS 2019 · similarity 0.91)
- BEDLAM: A Synthetic Dataset of Bodies Exhibiting Detailed Lifelike Animated Motion (CVPR 2023 · similarity 0.88)
- Capturing Humans in Motion: Temporal-Attentive 3D Human Pose and Shape Estimation From Monocular Video (CVPR 2022 · similarity 0.88)
- BEDLAM2.0: Synthetic humans and cameras in motion (NeurIPS 2025 · similarity 0.87)
- HumanVid: Demystifying Training Data for Camera-controllable Human Image Animation (NeurIPS 2024 · similarity 0.86)
- Motions as Queries: One-Stage Multi-Person Holistic Human Motion Capture (CVPR 2025 · similarity 0.86)
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.