Reproduce “PIE-NeRF: Physics-based Interactive Elastodynamics with NeRF” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yutao Feng, Yintong Shang, Xuan Li, Tianjia Shao, Chenfanfu Jiang, Yin Yang
PIE-NeRF introduces a framework for interactive, physics-based elastodynamic simulation of objects represented by Neural Radiance Fields. It enables physically accurate, nonlinear deformations at interactive rates by leveraging a meshless approach with quadratic Generalized Moving Least Squares (Q-GMLS) and a precise quadratic ray-warping scheme, preserving high visual quality.
Key insights
- Adaptive Poisson disk sampling combined with Q-GMLS kernels allows for effective meshless discretization of implicit NeRFs, enabling physics-based simulation without explicit mesh generation.
- Quadratic Generalized Moving Least Squares (Q-GMLS) is essential for robustly capturing nonlinear dynamics and large deformations, particularly for codimensional shapes prone to locking artifacts.
- Employing a quadratic ray-warping scheme, which leverages the Q-GMLS displacement field, is crucial for maintaining the original NeRF's visual quality even with substantial physical deformations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.