Research map / Video / Video generation
Combo: Co-speech holistic 3D human motion generation and efficient customizable adaptation in harmony
Chao Xu, Mingze Sun, Zhi-Qi Cheng, Fei Wang, Yang Liu, Baigui Sun, Ruqi Huang, Alexander G. Hauptmann
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 |
|---|---|---|---|
| use unified-plus-divided layers (l=7, j=1) instead of fully divided (ablation) | DU-Trans · Scott neutral | FMD: 1.265 -> 1.190 | Improved |
| add one divided layer to a fully unified stack (ablation) | DU-Trans · Scott neutral | FGD: 0.588 -> 0.579 | Improved |
| add face and body reconstruction losses (ablation) | DU-Trans · Scott neutral | FGD: 27.19 -> 0.579 | Improved |
| use loss weights 0.5 instead of 1 (ablation) | DU-Trans · Scott neutral | FMD: 1.412 -> 1.190 | Improved |
| use bi-directional flow with l=3 instead of l=1 (ablation) | DU-Trans · Scott neutral | FMD: 1.198 -> 1.098 | Improved |
| use bi-directional rather than unidirectional flow (ablation) | DU-Trans · Scott neutral | FMD: 1.126 -> 1.098 (Uni-BF vs l=3) | Improved |
| use more bi-flow layers (l=2,3,4) (ablation) | DU-Trans · Scott neutral | FMD: 1.098 -> 1.105 | No clear effect |
| place adapter in parallel instead of serial (ablation) | X-Adapter · Scott emotional | FMD: 1.332 -> 1.128 | Improved |
| add the condition input x to the adapter (ablation) | X-Adapter · Scott emotional | FMD: 1.236 -> 1.128 | Improved |
| use dynamic scale instead of fixed scalar 1.0 (ablation) | X-Adapter · Scott emotional | MSE: 7.044 -> 5.015 | Improved |
| use adapter rank 128 instead of prefix tuning (ablation) | X-Adapter · Scott emotional | FMD: 2.234 -> 1.128 | Improved |
| use adapter instead of LoRA-r64 (ablation) | X-Adapter · Scott emotional | FMD: 1.205 -> 1.128 | Improved |
Closest papers by meaning
- Towards Variable and Coordinated Holistic Co-Speech Motion Generation (CVPR 2024 · similarity 0.90)
- DiffSHEG: A Diffusion-Based Approach for Real-Time Speech-driven Holistic 3D Expression and Gesture Generation (CVPR 2024 · similarity 0.88)
- Emotional Speech-driven 3D Body Animation via Disentangled Latent Diffusion (CVPR 2024 · similarity 0.88)
- Co$^{\mathbf{3}}$Gesture: Towards Coherent Concurrent Co-speech 3D Gesture Generation with Interactive Diffusion (ICLR 2025 · similarity 0.88)
- The Language of Motion: Unifying Verbal and Non-verbal Language of 3D Human Motion (CVPR 2025 · similarity 0.88)
- EMAGE: Towards Unified Holistic Co-Speech Gesture Generation via Expressive Masked Audio Gesture Modeling (CVPR 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.
