Research map / 3D vision / Human pose estimation
PoseSyn: Synthesizing Diverse 3D Pose Data from In-the-Wild 2D Data
ChangHee Yang, Hyeonseop Song, Seokhun Choi, Seungwoo Lee, Jaechul Kim, Hoseok Do
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 |
|---|---|---|---|
| render images with human animation model instead of NeRF (ablation) | PoseSyn (3DCrowdNet) · 3DPW/EMDB/CMU/HuMMan mean (3D) | MPJPE: 98.5 -> 97.5 | Improved |
| use EEM and MSM challenging-pose synthesis instead of PoseGen's GAN poses | Ours-N (3DCrowdNet) · 3DPW/EMDB/CMU/HuMMan mean (3D) | MPJPE: 99.7 -> 98.5 | Improved |
| add PoseSyn synthesized data | 3DCrowdNet · 3DPW/EMDB/CMU/HuMMan mean (3D) | MPJPE: 103.2 -> 97.5 | Improved |
| add PoseSyn synthesized data | Hybrik · 3DPW/EMDB/CMU/HuMMan mean (3D) | MPJPE: 121.2 -> 104.6 | Improved |
| add PoseSyn synthesized data | 4DHumans · 3DPW/EMDB/CMU/HuMMan mean (3D) | MPJPE: 106.8 -> 99.1 | Improved |
| add PoseSyn synthesized data | 3DCrowdNet · LSPET/JHMDB mean (2D) | PCKh@0.6: 75.0 -> 76.0 | Improved |
| add PoseSyn synthesized data | 4DHumans · LSPET/JHMDB mean (2D) | PCKh@0.6: 87.3 -> 87.7 | Improved |
| add EEM to select challenging poses (ablation) | PoseSyn (3DCrowdNet) · 3DPW/EMDB/CMU/HuMMan mean (3D) | MPJPE: 101.1 -> 97.5 | Improved |
| use initial motion representation in MSM (ablation) | PoseSyn (3DCrowdNet) · 3DPW/EMDB/CMU/HuMMan mean (3D) | MPJPE: 98.5 -> 97.5 | Improved |
| use initial motion representation in MSM (ablation) | MSM motion synthesis · EMDB top-100 hard samples | min PA-MPJPE (mm): 151.1 -> 140.8 | Improved |
| use TPE itself instead of 4DHumans in EEM (ablation) | PoseSyn (3DCrowdNet) · 3DPW/EMDB/CMU/HuMMan mean (3D) | MPJPE: 98.8 -> 97.5 | Improved |
| use pretrained TPE weights for fine-tuning instead of training from scratch (ablation) | PoseSyn (3DCrowdNet) · 3DPW/EMDB/CMU/HuMMan mean (3D) | MPJPE: 98.8 -> 97.5 | Improved |
Closest papers by meaning
- PoseAug: A Differentiable Pose Augmentation Framework for 3D Human Pose Estimation (CVPR 2021 · similarity 0.89)
- In the Wild Human Pose Estimation Using Explicit 2D Features and Intermediate 3D Representations (CVPR 2019 · similarity 0.89)
- AdaptPose: Cross-Dataset Adaptation for 3D Human Pose Estimation by Learnable Motion Generation (CVPR 2022 · similarity 0.88)
- 3D Human Pose Estimation in the Wild by Adversarial Learning (CVPR 2018 · similarity 0.88)
- PoseGen: Learning to Generate 3D Human Pose Dataset with NeRF (AAAI 2024 · similarity 0.87)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 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.
