.png){kind=icon}
.png){kind=icon}
Return
Physics-guided human interaction generation via motion diffusion model
.png){kind=icon}
PRE
AI
DOI：10.1016/j.cviu.2025.104470.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Novel Framework: We present PhyInter, a denoising diffusion model guided by physical laws, enabling the generation of physically-plausible human-human interactions. • Score Function-based Differential Equation: We design a stochastic differential equation as guidance for the denoising diffusion process, enhanced by a human-human interaction multi-attention mechanism. • State-of-the-Art Performance: Our method achieves state-of-the-art performance on the InterHuman dataset, outperforming previous approaches without requiring additional training data.
Keywords:
PhyInter
denoising diffusion model
physical laws
human-human interactions
multi-attention mechanism
Journal
IF:
3.5
Papers:
428
Citations:
7.3K
Organization
No organization information available