FastMesh:Efficient Artistic Mesh Generation via Component Decoupling

2025-08-26

AI Summary

Existing mesh generation methods serialize manifold meshes into redundant vertex sequences, resulting in excessively long sequences and low efficiency. To address this, we propose a decoupled vertex-and-facet generation framework: first, an autoregressive model efficiently generates vertex coordinates; second, a bidirectional Transformer directly predicts the facet adjacency matrix, eliminating redundant vertex encoding. Furthermore, we introduce a fidelity enhancement module and topology-aware post-processing to suppress invalid connections and optimize geometric layout. Our approach significantly reduces sequence redundancy—compressing sequence length to just 23% of the current most compact method—while accelerating generation by over 8×. Moreover, it achieves superior geometric fidelity and topological validity compared to state-of-the-art methods.

Technology Category

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

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Recent mesh generation approaches typically tokenize triangle meshes into sequences of tokens and train autoregressive models to generate these tokens sequentially. Despite substantial progress, such token sequences inevitably reuse vertices multiple times to fully represent manifold meshes, as each vertex is shared by multiple faces. This redundancy leads to excessively long token sequences and inefficient generation processes. In this paper, we propose an efficient framework that generates artistic meshes by treating vertices and faces separately, significantly reducing redundancy. We employ an autoregressive model solely for vertex generation, decreasing the token count to approximately 23% of that required by the most compact existing tokenizer. Next, we leverage a bidirectional transformer to complete the mesh in a single step by capturing inter-vertex relationships and constructing the adjacency matrix that defines the mesh faces. To further improve the generation quality, we introduce a fidelity enhancer to refine vertex positioning into more natural arrangements and propose a post-processing framework to remove undesirable edge connections. Experimental results show that our method achieves more than 8$ imes$ faster speed on mesh generation compared to state-of-the-art approaches, while producing higher mesh quality.

Problem

• Reduces vertex redundancy in mesh generation
• Separates vertex and face generation processes
• Improves efficiency and quality of artistic meshes

Innovation

• Decouples vertices and faces generation
• Uses bidirectional transformer for adjacency
• Introduces fidelity enhancer for refinement