Reproduce “LaGSplat: Inferring Physics-Governed Interactive Simulation from Monocular Video Using Latent Lagrangian Gaussian Splatting” 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
Louen Pottier
LaGSplat learns a compact state from video that drives a dissipative Lagrangian and a Gaussian Splatting renderer. A decoder’s motion map converts image-space pushes into generalized forces, but force scale remains uncalibrated and unobserved motion modes are not learned. In a rainbow rocker rollout, amplitude falls to 0% after 1,000 seconds, versus 53% for GaussianPrediction, limited to this test.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.