Research map / Video / Video generation
Human Motion Diffusion Model
Guy Tevet, Sigal Raab, Brian Gordon, Yoni Shafir, Daniel Cohen-or, Amit Haim Bermano
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 MDM instead of T2M | text-to-motion generation · HumanML3D | FID: 1.067^{\pm.002} -> 0.544^{\pm.044} | Improved |
| use MDM instead of T2M | text-to-motion generation · KIT | FID: 2.770^{\pm.109} -> 0.497^{\pm.021} | Improved |
| use MDM instead of T2M | text-to-motion generation · HumanML3D | R Precision (top 3): 0.740^{\pm.003} -> 0.611^{\pm.007} | Hurt |
| use MDM instead of T2M | text-to-motion generation · KIT | R Precision (top 3): 0.693^{\pm.007} -> 0.396^{\pm.004} | Hurt |
| add foot contact loss (ablation) | MDM action-to-motion · HumanAct12 | FID: 0.080^{\pm.000} -> 0.100^{\pm.000} | Hurt |
| add foot contact loss (ablation) | MDM action-to-motion · HumanAct12 | Accuracy: 0.990^{\pm.000} -> 0.990^{\pm.000} | No clear effect |
| add foot contact loss (ablation) | MDM action-to-motion · UESTC | Accuracy: 0.960^{\pm.000} -> 0.950^{\pm.000} | Hurt |
| add foot contact loss (ablation) | MDM action-to-motion · UESTC | FID_test: 13.08^{\pm2.32} -> 12.81^{\pm1.46} | Improved |
| replace the encoder-only transformer with a GRU (ablation) | MDM text-to-motion · HumanML3D | FID: 0.544^{\pm.044} -> 4.569^{\pm.150} | Hurt |
| use a transformer decoder instead of the encoder-only backbone (ablation) | MDM text-to-motion · HumanML3D | FID: 0.544^{\pm.044} -> 0.767^{\pm.085} | Hurt |
| add an input token to the transformer decoder (ablation) | MDM decoder text-to-motion variant · HumanML3D | FID: 0.767^{\pm.085} -> 0.567^{\pm.051} | Improved |
| use MDM instead of MoDi | unconstrained motion generation · HumanAct12 | FID: 13.03 -> 31.92 | Hurt |
Closest papers by meaning
- Human Motion Diffusion as a Generative Prior (ICLR 2024 · similarity 0.92)
- Rethinking Diffusion for Text-Driven Human Motion Generation: Redundant Representations, Evaluation, and Masked Autoregression (CVPR 2025 · similarity 0.91)
- AMD: Autoregressive Motion Diffusion (AAAI 2024 · similarity 0.91)
- MotionMix: Weakly-Supervised Diffusion for Controllable Motion Generation (AAAI 2024 · similarity 0.90)
- Move-in-2D: 2D-Conditioned Human Motion Generation (CVPR 2025 · similarity 0.90)
- Single Motion Diffusion (ICLR 2024 · similarity 0.89)
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.
