Research map / Video / Video generation
Act As You Wish: Fine-Grained Control of Motion Diffusion Model with Hierarchical Semantic Graphs
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 hierarchical semantic graphs and coarse-to-fine diffusion (ablation) | GraphMotion · HumanML3D test | FID: 0.418^{\pm.013} -> 0.116^{\pm.007} | Helped |
| add semantic graph decomposition (ablation) | one-stage motion diffusion · HumanML3D test | FID: 0.418^{\pm.013} -> 0.281^{\pm.011} | Helped |
| add graph reasoning (ablation) | semantic graph motion diffusion · HumanML3D test | FID: 0.281^{\pm.011} -> 0.212^{\pm.015} | Helped |
| add coarse-to-fine diffusion (ablation) | semantic graph motion diffusion · HumanML3D test | FID: 0.281^{\pm.011} -> 0.196^{\pm.012} | Helped |
| add action and specific diffusion levels (ablation) | motion-level-only GraphMotion · HumanML3D test | FID: 0.217^{\pm.009} -> 0.116^{\pm.007} | Helped |
| add specific diffusion level (ablation) | motion-and-action-level GraphMotion · HumanML3D test | FID: 0.150^{\pm.011} -> 0.116^{\pm.007} | Helped |
| increase total steps from 50 to 150 (ablation) | GraphMotion · HumanML3D test | FID: 0.118^{\pm.008} -> 0.116^{\pm.007} | Helped |
| allocate more of 50 steps to the specific level (ablation) | GraphMotion · HumanML3D test | R-Precision Top-1: 0.489^{\pm.003} -> 0.496^{\pm.003} | Helped |
| use GraphMotion instead of MLD under verb/action-name masking | text-to-motion generation · HumanML3D test, masked verbs and action names | FID: 3.492 -> 1.826 | Helped |
| increase motion VAE token size from 2 to 8 (ablation) | motion VAE · HumanML3D test motion reconstruction | FID: 1.906 -> 0.019 | Helped |
| increase motion VAE token size from 2 to 8 (ablation) | motion VAE · KIT test motion reconstruction | FID: 1.196 -> 0.361 | Helped |
Closest papers by meaning
- Enhanced Fine-Grained Motion Diffusion for Text-Driven Human Motion Synthesis (AAAI 2024 · similarity 0.90)
- DartControl: A Diffusion-Based Autoregressive Motion Model for Real-Time Text-Driven Motion Control (ICLR 2025 · similarity 0.89)
- OMG: Towards Open-vocabulary Motion Generation via Mixture of Controllers (CVPR 2024 · similarity 0.89)
- AMD: Autoregressive Motion Diffusion (AAAI 2024 · similarity 0.88)
- Towards Detailed Text-to-Motion Synthesis via Basic-to-Advanced Hierarchical Diffusion Model (AAAI 2024 · similarity 0.88)
- MMM: Generative Masked Motion Model (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.