Physically grounded models improve wind motion prediction on trees
Wind on Trees: Testing Physical Grounding in Dynamic 4D Gaussian Splatting
Computer Vision and Pattern Recognition
Summary
Reconstructing how trees move in the wind from a single camera view is very hard because the exact motion can’t be seen clearly and the trees are always moving. The authors tested a new way to model tree motion, using physics-based rules like springs and damping to represent how parts of a tree vibrate in the wind. They showed this model predicts tree motion better when facing new wind conditions but doesn’t perfectly recover the exact physical parameters. This means the approach captures some real physical effects but is not yet fully accurate in learning all physical details.
What this means in practice
- •For computer graphics developers: Generate more realistic tree animations under varying wind conditions by integrating physics-based motion models into rendering systems.
- •For augmented reality engineers: Improve dynamic vegetation rendering in AR scenes by using physically grounded deformation models for more believable environmental effects.
Tested on simulated data.
Authors
Weiying Chen, Edmond Lou
Abstract
Monocular reconstruction of wind-driven vegetation is severely underconstrained: motion along the viewing direction is largely unobservable, a moving canopy offers few reliable correspondences, and nearly the entire scene is dynamic, providing little static reference. Directly-learned deformation fields in 4D Gaussian Splatting therefore optimize photometric consistency rather than recover the motion that produced it. We replace that field with a physically parameterized deformation prior: one damped harmonic oscillator per rigid part, driven by the observed wind and integrated by differentiable RK4, supervised photometrically alone. To test whether such a prior is physically grounded rather than merely well fit, we build a controlled synthetic testbed of three procedurally generated trees spanning an order of magnitude in skeleton complexity, whose per-part natural frequency follows from its own geometry and whose damping ratio is a fixed constant, both held out of training. On it, we measure held-out views, temporal extrapolation, zero-shot transfer to unseen wind speeds, and recovery of the physical parameters themselves. The prior costs appearance fidelity on in-distribution views and extrapolates markedly better outside the training window and the training wind, while parameter recovery is far weaker than it first appears: frequency recovery survives an untrained null control on only the sparsest of the three trees, and damping is not recovered at all.