Reproduce “SurfPhase: 3D Interfacial Dynamics in Two-Phase Flows from Sparse Videos” 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
Yue Gao, Hong-Xing Yu, Sanghyeon Chang, Qianxi Fu, Bo Zhu, Yoonjin Won, Juan Carlos Niebles, Jiajun Wu
A novel computational framework, SurfPhase, reconstructs the 3D geometry and velocity of liquid-vapor interfaces in two-phase flows using sparse-view videos. It integrates dynamic Gaussian surfels with signed distance function constraints and generative priors from video diffusion models to overcome limitations of sparse data and capture complex interfacial dynamics.
Key insights
- Integrating a Signed Distance Function (SDF) formulation with dynamic Gaussian surfels enables accurate reconstruction of sharp, topologically changing liquid-vapor interfaces from limited views.
- Leveraging video diffusion models to generate realistic novel-view videos effectively addresses the ill-posed nature of sparse-view 3D reconstruction by providing crucial generative priors.
- Physically informed, bubble-guided initialization of surfel positions is essential for establishing robust temporal correspondences and accurately estimating 3D interfacial velocities in highly dynamic two-phase flows.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.