Neural Mesh Flow: 3D Manifold Mesh Generation via Diffeomorphic Flows
This addresses a practical issue in computer graphics and 3D modeling where manifold meshes are required for realistic interactions in virtual environments, though it is incremental as it builds on existing shape auto-encoder and NODE techniques.
The paper tackles the problem of generating 3D meshes that are manifold (topologically correct) for applications like rendering and simulation, where prior methods produced accurate but non-manifold meshes. The proposed Neural Mesh Flow (NMF) generates two-manifold meshes for genus-0 shapes and demonstrates applicability in tasks such as single-view reconstruction and texture mapping.
Meshes are important representations of physical 3D entities in the virtual world. Applications like rendering, simulations and 3D printing require meshes to be manifold so that they can interact with the world like the real objects they represent. Prior methods generate meshes with great geometric accuracy but poor manifoldness. In this work, we propose Neural Mesh Flow (NMF) to generate two-manifold meshes for genus-0 shapes. Specifically, NMF is a shape auto-encoder consisting of several Neural Ordinary Differential Equation (NODE)[1] blocks that learn accurate mesh geometry by progressively deforming a spherical mesh. Training NMF is simpler compared to state-of-the-art methods since it does not require any explicit mesh-based regularization. Our experiments demonstrate that NMF facilitates several applications such as single-view mesh reconstruction, global shape parameterization, texture mapping, shape deformation and correspondence. Importantly, we demonstrate that manifold meshes generated using NMF are better-suited for physically-based rendering and simulation. Code and data are released.