Toronto AI Lab
NVIDIA Research
Meshes are ubiquitous in visual computing and simulation, yet most existing machine learning techniques represent meshes only indirectly, e.g. as the level set of a scalar field or deformation of a template, or as a disordered triangle soup lacking local structure. This work presents a scheme to directly generate manifold, polygonal meshes of complex connectivity as the output of a neural network. Our key innovation is to define a continuous latent connectivity space at each mesh vertex, which implies the discrete mesh. In particular, our vertex embeddings generate cyclic neighbor relationships in a halfedge mesh representation, which gives a guarantee of edge-manifoldness and the ability to represent general polygonal meshes. This representation is well-suited to machine learning and stochastic optimization, without restriction on connectivity or topology. We first explore the basic properties of this representation, then use it to fit distributions of meshes from large datasets. The resulting models generate diverse meshes with tessellation structure learned from the dataset population, with concise details and high- quality mesh elements. In applications, this approach not only yields high- quality outputs from generative models, but also enables directly learning challenging geometry processing tasks such as mesh repair.
Overview of the proposed representation.
Conditioned on the same geometry, our model can generate different styles of meshes depending to the distribution it was trained on. Each row denotes a style of mesh, for which we construct a dataset of meshed primitive surfaces and fit our model.
Trained on ABC dataset, our model generate high-quality meshes with vertices and edges align accurately with sharp features, highlighting the advantage of directly generating meshes as the output representation.
We further evaluate our model on ShapeNet dataset. Our method generates sharp and compact polygonal meshes that match with the input condition and are guaranteed to be manifold.
Envisioned workflow for interactive mesh repair.
A visual comparison of mesh repair methods.
@inproceedings{spacemesh2024,
author = {Tianchang Shen and Zhaoshuo Li and Marc Law and Matan Atzmon and Sanja Fidler
and James Lucas and Jun Gao and Nicholas Sharp},
title = {SpaceMesh: A Continuous Representation for Learning Manifold Surface Meshes},
booktitle = {SIGGRAPH Asia 2024 Conference Papers (SA Conference Papers '24)},
year = {2024},
location = {Tokyo, Japan},
date = {December 3-6, 2024},
publisher = {ACM},
address = {New York, NY, USA},
pages = {11},
doi = {10.1145/3680528.3687634},
url = {https://doi.org/10.1145/3680528.3687634}
}
SpaceMesh: A Continuous Representation for Learning Manifold Surface Meshes
Tianchang Shen, Zhaoshuo Li, Marc Law, Matan Atzmon, Sanja Fidler, James Lucas, Jun Gao, Nicholas Sharp
|
|