Abstract:This paper introduces a new learning-based method, NASM, for anisotropic surface meshing. Our key idea is to propose a graph neural network to embed an input mesh into a high-dimensional (high-d) Euclidean embedding space to preserve curvature-based anisotropic metric by using a dot product loss between high-d edge vectors. This can dramatically reduce the computational time and increase the scalability. Then, we propose a novel feature-sensitive remeshing on the generated high-d embedding to automatically capture sharp geometric features. We define a high-d normal metric, and then derive an automatic differentiation on a high-d centroidal Voronoi tessellation (CVT) optimization with the normal metric to simultaneously preserve geometric features and curvature anisotropy that exhibit in the original 3D shapes. To our knowledge, this is the first time that a deep learning framework and a large dataset are proposed to construct a high-d Euclidean embedding space for 3D anisotropic surface meshing. Experimental results are evaluated and compared with the state-of-the-art in anisotropic surface meshing on a large number of surface models from Thingi10K dataset as well as tested on extensive unseen 3D shapes from Multi-Garment Network dataset and FAUST human dataset.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is **how to generate high - quality anisotropic surface meshes efficiently and robustly**, especially without providing pre - computed metric fields and automatically preserving geometric features (such as sharp and weak features). Specifically, the paper focuses on the following key challenges: 1. **Efficient high - dimensional embedding**: Traditional methods require a large amount of computing time when generating anisotropic meshes, especially when dealing with complex shapes. Therefore, the paper proposes a method based on graph neural networks (GNN) to embed the input mesh into a high - dimensional Euclidean space to maintain the anisotropic metric of the curvature basis. This method can significantly reduce the computing time and improve the scalability. 2. **Automatically generate high - quality anisotropic meshes**: Existing anisotropic mesh generation methods usually require users to manually mark feature edges and corners and constrain these features during the optimization process. This is not only cumbersome but also error - prone. The paper proposes a new feature - sensitive remeshing method. After generating the high - dimensional embedding, it automatically captures geometric features, thereby generating high - quality anisotropic meshes. 3. **Application of large - scale datasets**: To verify the effectiveness and generalization ability of the method, the paper constructs a large - scale dataset containing more than 800 mesh models, covering a variety of geometric shapes and topological structures. Through experiments on these datasets, the superior performance of the proposed method on different shapes is proved. ### Main contributions of the paper - **Developed a scalable computing paradigm**: High - quality anisotropic surface meshes can be generated only by inputting any mesh without providing curvature metrics. - **Designed an efficient GNN method**: The input mesh is embedded into a high - dimensional Euclidean space through high - dimensional dot - product loss, which is about 1500 times faster than traditional methods. - **Defined a high - dimensional normal metric CVT optimization method**: Feature - sensitive anisotropic surface meshes are calculated through automatic differentiation without users manually marking features. - **Constructed a large - scale dataset**: For anisotropic surface mesh generation and processing, including more than 800 mesh models from Thingi10K, Multi - Garment Network and FAUST datasets. ### Summary By combining deep learning and high - dimensional embedding techniques, the paper proposes a brand - new anisotropic surface mesh generation method NASM, which solves the problems of low computational efficiency, dependence on manual feature annotation and difficulty in handling large - scale datasets in existing methods. This method not only improves the quality and efficiency of mesh generation but also provides strong support for subsequent geometric modeling, physical simulation and mechanical engineering fields.

NASM: Neural Anisotropic Surface Meshing

MeshingNet: A New Mesh Generation Method based on Deep Learning

MeshPointNet: 3D Surface Classification Using Graph Neural Networks and Conformal Predictions on Mesh-Based Representations

High-Fidelity Mask-free Neural Surface Reconstruction for Virtual Reality

MeshNet: Mesh Neural Network for 3D Shape Representation

Developing a novel structured mesh generation method based on deep neural networks

DMNet: Delaunay Meshing Network for 3D Shape Representation

SpaceMesh: A Continuous Representation for Learning Manifold Surface Meshes

HR-NeuS: Recovering High-Frequency Surface Geometry via Neural Implicit Surfaces

Mesh Neural Networks Based on Dual Graph Pyramids

NeuDA: Neural Deformable Anchor for High-Fidelity Implicit Surface Reconstruction

NeAT: Learning Neural Implicit Surfaces with Arbitrary Topologies from Multi-view Images

3DMeshNet: A Three-Dimensional Differential Neural Network for Structured Mesh Generation

Neural Volumetric Mesh Generator

AniSDF: Fused-Granularity Neural Surfaces with Anisotropic Encoding for High-Fidelity 3D Reconstruction

Neural Progressive Meshes

MeshFormer: High-Quality Mesh Generation with 3D-Guided Reconstruction Model

Proposing an intelligent mesh smoothing method with graph neural networks

Deep Spectral Meshes: Multi-Frequency Facial Mesh Processing with Graph Neural Networks

PAI-GCN: Permutable Anisotropic Graph Convolutional Networks for 3D Shape Representation Learning.