Research map / 3D vision / Human pose estimation
Sim2real transfer learning for 3D human pose estimation: motion to the rescue
Read the paperCheck 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 |
|---|---|---|---|
| Use Flow + Keypoints instead of RGB Only | Motion HMR trained on synthetic SURREAL data · 3DPW test | PA-MPJPE: 105.6 -> 74.7 | Helped |
| Use Keypoints Only instead of RGB Only | Motion HMR trained on synthetic SURREAL data · 3DPW test | PA-MPJPE: 105.6 -> 77.6 | Helped |
| Add optical flow to Keypoints Only | Motion HMR trained on synthetic SURREAL data · 3DPW test | PA-MPJPE: 77.6 -> 74.7 | Helped |
| Add RGB to Keypoints Only | Motion HMR trained on synthetic SURREAL data · 3DPW test | PA-MPJPE: 77.6 -> 82.4 | Hurt |
| Add DANN to RGB | Motion HMR trained on synthetic SURREAL data · 3DPW test | PA-MPJPE: 105.6 -> 103.0 (107.5) | No clear effect |
| Add realistic moving backgrounds to synthetic training videos (ablation) | Motion HMR with Flow + Keypoints · 3DPW test | PA-MPJPE: 88.9 -> 80.3 | Helped |
| Track background camera motion in synthetic videos (ablation) | Motion HMR with Flow + Keypoints · 3DPW test | PA-MPJPE: 80.3 -> 77.2 | Helped |
| Add synthetic occlusions (ablation) | Motion HMR with Flow + Keypoints · 3DPW test | PA-MPJPE: 77.2 -> 74.7 | Helped |
| Use 31-frame clips instead of 8-frame clips (ablation) | Motion HMR with Flow + Keypoints · 3DPW test | PA-MPJPE: 79.0 -> 74.7 | Helped |
| Use 31-frame clips instead of 16-frame clips (ablation) | Motion HMR with Flow + Keypoints · 3DPW test | PA-MPJPE: 77.8 -> 74.7 | Helped |
| Extend clips from 31 to 56 frames (ablation) | Motion HMR with Flow + Keypoints · 3DPW test | PA-MPJPE: 74.7 -> 76.2 | Hurt |
| Use synthetic-only 3D pose training with Flow + Keypoints rather than real-data HMR | Motion HMR · 3DPW test | PA-MPJPE: 77.2 -> 74.7 | Helped |
Closest papers by meaning
- Learning From Synthetic Humans (CVPR 2017 · similarity 0.91)
- AdaptPose: Cross-Dataset Adaptation for 3D Human Pose Estimation by Learnable Motion Generation (CVPR 2022 · similarity 0.90)
- VideoMDM: Towards 3D Human Motion Generation From 2D Supervision (NeurIPS 2026 · similarity 0.89)
- Progressive Multi-View Human Mesh Recovery with Self-Supervision (AAAI 2023 · similarity 0.89)
- Direct Prediction of 3D Body Poses From Motion Compensated Sequences (CVPR 2016 · similarity 0.89)
- TokenHMR: Advancing Human Mesh Recovery with a Tokenized Pose Representation (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.