Research map / 3D vision / Human pose estimation
A Unified 3D Human Motion Synthesis Model via Conditional Variational Auto-Encoder
Yujun Cai, Yiwei Wang, Yiheng Zhu, Tat-Jen Cham, Jianfei Cai, Junsong Yuan, Jun Liu, Chuanxia Zheng, Sijie Yan, Henghui Ding, Xiaohui Shen, Ding Liu and 1 more
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 Action-Adaptive Modulation instead of concatenating action label (ablation) | Ours w/ action · Human3.6M action classification | average classification accuracy: 0.65 -> 0.91 | Improved |
| add action labels (concatenation, no AAM) (ablation) | Ours w/o action · Human3.6M action classification | average classification accuracy: 0.45 -> 0.65 | Improved |
| add cross-attention mechanism (ablation) | Ours w/ action · Human3.6M | FIDg: 71.3 -> 62.5 | Improved |
| add cross-attention mechanism (ablation) | Ours w/ action · Human3.6M | MPJPE: 104.4 -> 97.3 | Improved |
| add cross-attention mechanism (ablation) | Ours w/ action · Human3.6M | IS: 6.43 -> 6.87 | Improved |
| add cross-attention mechanism (ablation) | Ours w/ action · Human3.6M | Diversity: 0.18 -> 0.18 | No clear effect |
| add action labels with AAM and classifier | Ours w/o action · Human3.6M | FIDg: 88.2 -> 62.5 | Improved |
| add action labels with AAM and classifier | Ours w/o action · Human3.6M | MPJPE: 115.6 -> 97.3 | Improved |
| add action labels with AAM and classifier | Ours w/o action · Human3.6M | Diversity: 0.26 -> 0.19 | Hurt |
| add action labels with AAM and classifier | Ours w/o action · CMU-Mocap | FIDa: 162.3 -> 97.7 | Improved |
| add action labels with AAM and classifier | Ours w/o action · CMU-Mocap | Diversity: 0.08 -> 0.05 | Hurt |
| use CVAE ours w/o action instead of CSGN | CSGN · Human3.6M | FIDg: 136.8 -> 88.2 | Improved |
Closest papers by meaning
- Learning Uncoupled-Modulation CVAE for 3D Action-Conditioned Human Motion Synthesis (ECCV 2022 · similarity 0.93)
- Action-Conditioned 3D Human Motion Synthesis With Transformer VAE (ICCV 2021 · similarity 0.90)
- Contextually Plausible and Diverse 3D Human Motion Prediction (ICCV 2021 · similarity 0.90)
- Seamless Human Motion Composition with Blended Positional Encodings (CVPR 2024 · similarity 0.90)
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.89)
- A Unified Masked Autoencoder with Patchified Skeletons for Motion Synthesis (AAAI 2024 · similarity 0.88)
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.
