VertexRegen: Mesh Generation with Continuous Level of Detail

2025-08-12

AI Summary

Existing autoregressive mesh generation methods adopt a “part-to-whole” modeling paradigm, yielding intermediate outputs that lack geometric integrity and semantic consistency. To address this, we propose a progressive mesh generation framework based on learnable vertex splitting—the first generative approach to explicitly incorporate edge splitting, the inverse operation of edge collapse, enabling continuous and controllable reconstruction from coarse topology to fine-grained geometry. Our method supports anytime termination: at any generation step, it produces topologically valid meshes with adjustable levels of geometric detail. This yields true anytime generation capability without post-hoc refinement. Experiments demonstrate state-of-the-art mesh quality in terms of fidelity and diversity, while significantly improving flexibility in detail control and geometric plausibility throughout the generation process.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Segmentation
• Natural Language Processing: Generation

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

We introduce VertexRegen, a novel mesh generation framework that enables generation at a continuous level of detail. Existing autoregressive methods generate meshes in a partial-to-complete manner and thus intermediate steps of generation represent incomplete structures. VertexRegen takes inspiration from progressive meshes and reformulates the process as the reversal of edge collapse, i.e. vertex split, learned through a generative model. Experimental results demonstrate that VertexRegen produces meshes of comparable quality to state-of-the-art methods while uniquely offering anytime generation with the flexibility to halt at any step to yield valid meshes with varying levels of detail.

Problem

• Generates meshes with continuous detail levels
• Replaces partial-to-complete mesh generation methods
• Enables flexible stopping for valid multi-detail meshes

Innovation

• Continuous level of detail mesh generation
• Reversal of edge collapse via vertex split
• Anytime generation with flexible stopping