LATO.2: Factorized 3D Mesh Generation with Vertex and Topology Flow
2026-07-12 • Graphics
GraphicsComputer Vision and Pattern Recognition
AI summaryⓘ
The authors introduce LATO.2, a new method for creating 3D meshes by breaking the process into two steps: first generating vertex positions, then determining how those vertices connect. This separation helps avoid problems seen in previous methods that mixed geometry and connectivity together, leading to errors like drifting points or broken shapes. By structuring the process around a shared voxel grid and using special autoencoders, their approach achieves higher detail and allows easy local editing without retraining. Their experiments show that LATO.2 produces more accurate and well-connected 3D meshes than earlier techniques.
flow matchinglatent representationmesh generationvertex geometryconnectivityVAE (variational autoencoder)voxel scaffoldtopology-aware3D meshlatent space factorization
Authors
Hang Long, Tianhao Zhao, Junkai Lin, Youjia Zhang, Huipeng Guo, Rendong Liang, Jiale Xu, Jozef Hladký, Matthias Nießner, Wei Yang
Abstract
Flow matching over carefully designed latent representations has recently emerged as a powerful paradigm for topology-aware mesh generation. Existing approaches, however, model vertices and connectivity jointly in a joint latent space, entangling continuous vertex geometry with discrete combinatorial structure; this complicates flow learning and manifests as drifting vertices and broken surfaces. We present LATO.2, a factorized flow matching framework that decomposes mesh generation into a vertex flow followed by a connectivity flow conditioned on the realized vertices, with both stages anchored to a shared coarse voxel scaffold. Dedicated VAEs underpin the two stages, recovering vertices at sub-voxel precision and embedding discrete connectivity into a continuous latent space. We demonstrate two advantages unique to this factorization: (i) part-wise generation, in which the scaffold is partitioned and each part synthesized at full latent capacity, yielding substantially higher-resolution meshes than a monolithic latent permits; and (ii) topology-adaptive editing, in which manipulating first-stage vertices induces the corresponding connectivity without re-optimization. Experiments show that LATO.2 surpasses state-of-the-art topology-aware mesh generators in geometric fidelity and connectivity quality.