Quad mesh generation improves with singularity aware serialization
Q-MAGIC: Quad-only Mesh Autoregressive Generation via Singularity-Aware Circular Relative Offset Tokenization
Computer Vision and Pattern RecognitionGraphics
Summary
Making 3D shapes from four-sided faces is tricky because the faces need to fit together properly. The authors designed a new way to break down and represent these four-sided faces so a computer can learn to make them step-by-step. Their method pays special attention to certain important points in the shape to guide the process. Using this approach, their model creates better 3D shapes with quads more efficiently than earlier methods.
What this means in practice
- •For 3d artists and modelers: Generate high-quality quad-only 3D meshes automatically for use in design and animation workflows.
- •For computer graphics software developers: Integrate an autoregressive quad mesh generation method to improve procedural content creation tools.
Authors
Yuming Zhao, Zezeng Li, Weimin Wang, Xiang Wu, Na Lei
Abstract
Autoregressive models have proven effective for triangular mesh generation by transforming it into predicting the next face, converting the non-differentiable connectivity prediction into a sequence of token predictions. However, generating quad-only meshes remains challenging because quad faces impose stricter topological constraints than triangles. The model must not only serialize quad faces into a valid token sequence, but also support unambiguous connectivity reconstruction during de-serialization. To address this, we propose Q-MAGIC, a novel autoregressive model for quad mesh generation. We first introduce a tailored serialization scheme for quad meshes, which requires at least two new vertices for the next face and is less flexible than triangles due to strict wiring constraints. To meet these requirements and reduce serialization length, we propose a singularity-aware serialization that prioritizes wiring from singular points, leading to more coherent traversals. Additionally, we propose a circular relative offset tokenization that encodes geometry mainly via local relative offsets. Built upon this representation, an autoregressive Transformer learns the underlying distribution of quad meshes via next-token prediction, enabling effective quad mesh generation. Extensive experiments demonstrate that Q-MAGIC produces more compact sequences while consistently improving geometric fidelity and mesh quality compared to state-of-the-art methods.