CraftsMan:
High-fidelity Mesh Generation with 3D Native Generation and Interactive Geometry Refiner

Weiyu Li* 1,2,  

Jiarui Liu*

1,2,   Rui Chen 1,   Yixun Liang 2,3,   Xuelin Chen 4,   Ping Tan 1,2,   Xiaoxiao Long† 5

1 HKUST

,

2 LightIllusions

,

3 HKUST(GZ)

,

4 Tencent AI Lab

,

5 HKU

Abstract

We present a novel generative 3D modeling system, coined CraftsMan, which can generate high-fidelity 3D geometries with highly varied shapes, regular mesh topologies and detailed surfaces, and notably, allows for refining the geometry in an interactive manner. Despite the significant advancements in 3D generation, existing methods still struggle with lengthy optimization processes, irregular mesh topologies, noisy surfaces, and difficulties in accommodating user edits, consequently impeding their widespread adoption and implementation in 3D modeling software. Our work is inspired by the craftsman, who usually roughs out the holistic figure of the work first and elaborate the surface details subsequently. Specifically, we employ a 3D native diffusion model, which operates on latent space learned from latent set-based 3D representations, to generate coarse geometries with regular mesh topology in seconds. In particular, this process takes as input a text prompt or a reference image, and leverages a powerful multi-view (MV) diffusion model to generate multiple views of the coarse geometry, which are fed into our MV-conditioned 3D diffusion model for generating the 3D geometry, significantly improving robustness and generalizability. Following that, a normal-based geometry refiner is used to significantly enhance the surface details. This refinement can be performed automatically, or inter- actively with user-supplied edits. Extensive experiments demonstrate that our method achieves high efficiency in producing superior-quality 3D assets compared to existing methods.

Image Conditioned 3D Native Diffusion Results

Reference Image

Output Coarse Mesh

Reference Image

Output Coarse Mesh

More Results

Text Conditioned 3D Native Diffusion Results

A cute panda

Leaning Tower of Pisa

Viking axe, fantasy, weapon

A cute rabbit

More Results

Normal Brush Results

More Results

Method Overview

Our method first transforms the input single image or text prompt into multi-view images using a multi-view diffusion model. These generated multi-view images are then fed into a native 3D diffusion model as conditions to produce a coarse mesh with regular topology. Finally, a surface normal-based refinement is employed to improve or edit the coarse geometry, enhancing it with intricate details. The refinement process features two key tools: an automatic global refinement and an interactive Magic Brush, which together enable efficient and controllable 3D modeling.

BibTeX

@article{li2024craftsman,
author    = {Weiyu Li and Jiarui Liu and Rui Chen and Yixun Liang and Xuelin Chen and Ping Tan and Xiaoxiao Long},
title     = {CraftsMan: High-fidelity Mesh Generation with 3D Native Generation and Interactive Geometry Refiner},
journal   = {arXiv preprint arXiv:2405.14979},
year      = {2024},
}