8000 GitHub - haoz19/PhysRig: Physics-based rigging with MPM for realistic character animation. ICCV 2025.
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

haoz19/PhysRig

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 

Repository files navigation

🔩 PhysRig: Differentiable Physics-Based Rigging for Realistic Articulated Object Modeling

Paper Project Page ICCV 2025

PhysRig is a differentiable, physics-based skinning and rigging framework that enables realistic deformation of articulated 3D objects. Unlike traditional methods like Linear Blend Skinning (LBS), PhysRig embeds skeletons into a deformable soft-body volume simulated using Material Point Method (MPM) — capturing the behavior of soft tissues, tails, ears, and other elastic structures in a physically plausible way.

🔧 From UIUC & Stability AI | 🦖 ICCV 2025 Accepted


🧠 Highlights

  • Differentiable Physics-Based Simulation
    Deformations are modeled using MPM and continuum mechanics, supporting gradient-based optimization of both motion and material properties.

  • Material Prototypes
    Learnable elastic parameter templates (Young’s modulus & Poisson’s ratio) provide expressive yet compact material modeling.

  • Driving Point Rigging
    Instead of bone transformations, PhysRig uses velocity-driven embedded points to induce dynamic shape changes.

  • Strong Performance
    Outperforms LBS on both user studies and geometric metrics (Chamfer Distance), with better convergence and physical realism.

  • Applications

    • Inverse Skinning (from motion to parameters)
    • Pose Transfer (e.g., human → jellyfish)
    • 4D Reconstruction & Animation

Code will be released to the public around ICCV

About

Physics-based rigging with MPM for realistic character animation. ICCV 2025.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published
0