MeshAnything V2: Artist-Created Mesh Generation With Adjacent Mesh Tokenization

TL;DR

MeshAnything V2 introduces a novel tokenization method that reduces sequence length and doubles face limits in autoregressive mesh generation, enabling more complex and efficient creation of artist-designed 3D meshes.

Contribution

The paper presents Adjacent Mesh Tokenization (AMT), a new approach that improves mesh sequence efficiency and model capacity in autoregressive mesh generation.

Findings

AMT reduces token sequence length by about half.

MeshAnything V2 doubles the face limit compared to previous models.

The model achieves superior performance in generating complex meshes.

Abstract

Meshes are the de facto 3D representation in the industry but are labor-intensive to produce. Recently, a line of research has focused on autoregressively generating meshes. This approach processes meshes into a sequence composed of vertices and then generates them vertex by vertex, similar to how a language model generates text. These methods have achieved some success but still struggle to generate complex meshes. One primary reason for this limitation is their inefficient tokenization methods. To address this issue, we introduce MeshAnything V2, an advanced mesh generation model designed to create Artist-Created Meshes that align precisely with specified shapes. A key innovation behind MeshAnything V2 is our novel Adjacent Mesh Tokenization (AMT) method. Unlike traditional approaches that represent each face using three vertices, AMT optimizes this by employing a single vertex wherever feasible, effectively reducing the token sequence length by about half on average. This not only streamlines the tokenization process but also results in more compact and well-structured sequences, enhancing the efficiency of mesh generation. With these improvements, MeshAnything V2 effectively doubles the face limit compared to previous models, delivering superior performance without increasing computational costs. We will make our code and models publicly available. Project Page: https://buaacyw.github.io/meshanything-v2/