MeshRipple: Structured Autoregressive Generation of Artist-Meshes

2025-12-08

AI Summary

Existing autoregressive mesh generators are memory-constrained, relying on truncated sequences and sliding-window inference, which impedes modeling of long-range geometric dependencies—resulting in holes and topological fragmentation in generated meshes. To address this, we propose MeshRipple, a ripple-expansion generation framework: (1) a frontier-aware BFS-based face serialization strategy ensures local connectivity; (2) a dilation-based prediction mechanism progressively extrapolates the mesh boundary; and (3) a hybrid attention architecture—combining sparse local attention with a global memory module—explicitly encodes long-range topological constraints. MeshRipple preserves the advantages of autoregressive modeling while significantly improving surface fidelity and topological integrity. It achieves state-of-the-art performance across multiple benchmarks and, for the first time, enables end-to-end, hole-free, and topologically consistent 3D mesh generation.

Technology Category

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

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Meshes serve as a primary representation for 3D assets. Autoregressive mesh generators serialize faces into sequences and train on truncated segments with sliding-window inference to cope with memory limits. However, this mismatch breaks long-range geometric dependencies, producing holes and fragmented components. To address this critical limitation, we introduce MeshRipple, which expands a mesh outward from an active generation frontier, akin to a ripple on a surface.MeshRipple rests on three key innovations: a frontier-aware BFS tokenization that aligns the generation order with surface topology; an expansive prediction strategy that maintains coherent, connected surface growth; and a sparse-attention global memory that provides an effectively unbounded receptive field to resolve long-range topological dependencies.This integrated design enables MeshRipple to generate meshes with high surface fidelity and topological completeness, outperforming strong recent baselines.

Problem

• Autoregressive mesh generation breaks long-range geometric dependencies
• Current methods produce holes and fragmented components in meshes
• MeshRipple resolves topological dependencies for coherent surface growth

Innovation

• Frontier-aware BFS tokenization aligns generation with topology
• Expansive prediction strategy ensures coherent connected surface growth
• Sparse-attention global memory provides unbounded receptive field