Search papers, labs, and topics across Lattice.
This paper introduces a Monte Carlo method for estimating the Dirichlet-to-Neumann (DtN) operator and its Steklov eigenmodes, addressing the limitations of intrinsic methods in geometry processing when applied to in-the-wild meshes. By leveraging volumetric constructions, the method effectively couples disconnected mesh components and demonstrates significant speed improvements over traditional boundary-element approaches, achieving robustness against poor mesh quality. The authors validate their approach by computing Steklov eigenspectra for approximately 450,000 shapes from the Objaverse dataset and integrating these operators into a neural network for large-scale 3D representation learning.
Monte Carlo methods can now compute Steklov spectra orders of magnitude faster while handling complex, disconnected geometries in real-world datasets.
Intrinsic methods fill the default toolbox for geometry processing on meshes. Intrinsic operators, in particular the Laplacian, underlie methods that require invariance to isometry and have hence been employed in many algorithms for shape analysis, learning, and editing. However, intrinsic methods are predicated on assumptions that quickly become brittle when working with in-the-wild geometry, where (i) mesh quality is not guaranteed, and (ii) many meshes are modeled with multiple connected components. In such settings, volumetric constructions are better-defined, since restrictions on surface topology can be relaxed. This paper presents a Monte Carlo method for estimating the Dirichlet-to-Neumann (DtN) operator -- a boundary-to-boundary volumetric operator -- and its associated Steklov eigenmodes. We build on recent developments in Monte Carlo geometry processing by casting this boundary operator itself as the subject of estimation. The DtN operator, defined through a volumetric stochastic process, is then generalized to the exterior domain, where it couples disconnected components through the surrounding ambient space. We show that our method is orders of magnitude faster than existing boundary-element approaches for computing Steklov spectra while remaining robust to poor triangulations, high-resolution meshes, and multi-component geometry. To demonstrate this scalability, we compute interior and exterior Steklov eigenspectra for approximately 450,000 shapes from the uncurated Objaverse dataset. We incorporate these operators into Steklov-CLIP, a mesh-based neural network that uses volumetric spectral operators for large-scale contrastive 3D representation learning. The resulting network learns semantically meaningful global and dense shape representations, illustrating that geometrically-principled volumetric operators can be made practical at the scale of modern 3D datasets.
MIT CSAIL