Research map / Video / Video generation
MotionDiffuse: Text-Driven Human Motion Generation With Diffusion Model
Mingyuan Zhang, Zhongang Cai, Liang Pan, Fangzhou Hong, Xinying Guo, Lei Yang, Ziwei Liu
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 pretrained CLIP text encoder (ablation) | MotionDiffuse with efficient attention · KIT-ML | R Precision Top-1: 0.136 -> 0.417 | Improved |
| add pretrained CLIP text encoder (ablation) | MotionDiffuse without efficient attention · KIT-ML | R Precision Top-1: 0.288 -> 0.357 | Improved |
| add efficient attention with CLIP (ablation) | MotionDiffuse · KIT-ML | R Precision Top-1: 0.357 -> 0.417 | Improved |
| add efficient attention without CLIP (ablation) | MotionDiffuse · KIT-ML | R Precision Top-1: 0.288 -> 0.136 | Hurt |
| raise latent dimension from 256 to 512 (ablation) | 8-layer MotionDiffuse · KIT-ML | R Precision Top-1: 0.198 -> 0.417 | Improved |
| raise latent dimension from 128 to 256 (ablation) | 8-layer MotionDiffuse · KIT-ML | R Precision Top-1: 0.025 -> 0.198 | Improved |
| increase layers from 4 to 8 (ablation) | MotionDiffuse dim 512 · KIT-ML | R Precision Top-1: 0.405 -> 0.417 | No clear effect |
| increase layers from 8 to 12 (ablation) | MotionDiffuse dim 512 · KIT-ML | R Precision Top-1: 0.417 -> 0.412 | No clear effect |
| increase layers from 4 to 8 (ablation) | MotionDiffuse dim 256 · KIT-ML | R Precision Top-1: 0.095 -> 0.198 | Improved |
| increase latent dimension from 128 to 512 (ablation) | 8-layer MotionDiffuse · HumanAct12 | FID: 0.22 -> 0.07 | Improved |
| increase layers from 4 to 8 (ablation) | MotionDiffuse dim 512 · HumanAct12 | FID: 0.09 -> 0.07 | Improved |
| use MotionDiffuse over Guo et al. | text-to-motion · HumanML3D | FID: 1.067 -> 0.630 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MoCoGAN — MoCoGAN: Decomposing Motion and Content for Video Generation (CVPR 2018) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Enhanced Fine-Grained Motion Diffusion for Text-Driven Human Motion Synthesis (AAAI 2024 · similarity 0.92)
- Human Motion Diffusion Model (ICLR 2023 · similarity 0.91)
- Human Motion Diffusion as a Generative Prior (ICLR 2024 · similarity 0.90)
- MotionMix: Weakly-Supervised Diffusion for Controllable Motion Generation (AAAI 2024 · similarity 0.90)
- AMD: Autoregressive Motion Diffusion (AAAI 2024 · similarity 0.90)
- MMM: Generative Masked Motion Model (CVPR 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.
