File size: 2,993 Bytes
7c70503 a91090e 7c70503 a91090e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 | ---
license: mit
language:
- en
pipeline_tag: image-to-3d
---
<div align="center">
<h1>LATO.2: Factorized 3D Mesh Generation with Vertex and Topology Flow</h1>
<p>
<a href="https://github.com/LoHhhha">Hang Long</a><sup>1,2,*</sup>
<a href="https://tianhaozhao668.github.io">Tianhao Zhao</a><sup>1,2,*</sup>
<a href="https://github.com/MayMhappy">Junkai Lin</a><sup>1,2</sup>
<a href="https://youjiazhang.github.io">Youjia Zhang</a><sup>1,2</sup>
Huipeng Guo<sup>1</sup><br>
<a href="https://github.com/PENGUINLIONG">Rendong Liang</a><sup>2</sup>
<a href="https://bluestyle97.github.io">Jiale Xu</a><sup>2</sup>
Jozef Hladký<sup>3</sup>
<a href="https://niessnerlab.org">Matthias Nießner</a><sup>4</sup>
<a href="https://yuanming.taichi.graphics">Yuanming Hu</a><sup>2</sup>
<a href="https://weiyang-hust.github.io">Wei Yang</a><sup>1,†</sup>
</p>
<p>
<sup>1</sup>Huazhong University of Science and Technology ·
<sup>2</sup>Meshy AI ·
<sup>3</sup>Independent Researcher ·
<sup>4</sup>Technical University of Munich
</p>
<p><sup>*</sup>Equal contribution · <sup>†</sup>Corresponding author</p>
<p><em>This work was completed during internships at Meshy AI.</em></p>
<p>
<a href="https://arxiv.org/abs/2607.10623">
<img src="https://img.shields.io/badge/arXiv-b31b1b.svg?logo=arXiv&logoColor=white" alt="arXiv">
</a>
<a href="https://github.com/LoHhhha/LATO.2">
<img src="https://img.shields.io/badge/GitHub-%23121011.svg?logo=github&logoColor=white" alt="GitHub">
</a>
</p>
</div>
## 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. |