Consistency^2: Consistent and Fast 3D Painting with Latent Consistency Models
Tianfu Wang, Anton Obukhov, Konrad Schindler
2024-06-18
Summary
This paper introduces Consistency^2, a new method for quickly and consistently painting 3D objects using a type of machine learning model called a Latent Consistency Model (LCM). This approach aims to improve the efficiency and quality of 3D painting.
What's the problem?
3D painting, which involves adding color and texture to 3D models, can be slow and complicated. Traditional methods often rely on time-consuming processes that require many steps to ensure the painted surfaces look good and consistent. Additionally, most techniques designed for 2D images don't easily translate to 3D painting, making it hard to improve the speed and quality of these methods.
What's the solution?
To solve these issues, the authors developed Consistency^2, which adapts LCMs for 3D painting. This method allows for faster generation of textures on 3D models while ensuring that the painted areas remain visually consistent. They tested their approach using a dataset called Objaverse and found that their method performed well in both speed and quality compared to existing techniques. The paper also provides an analysis of the strengths and weaknesses of their model.
Why it matters?
This research is important because it represents a significant advancement in the field of 3D content creation. By making it easier and faster to paint 3D objects, Consistency^2 can help artists and designers work more efficiently. This could lead to better tools for video games, movies, and other industries that rely on high-quality 3D graphics.
Abstract
Generative 3D Painting is among the top productivity boosters in high-resolution 3D asset management and recycling. Ever since text-to-image models became accessible for inference on consumer hardware, the performance of 3D Painting methods has consistently improved and is currently close to plateauing. At the core of most such models lies denoising diffusion in the latent space, an inherently time-consuming iterative process. Multiple techniques have been developed recently to accelerate generation and reduce sampling iterations by orders of magnitude. Designed for 2D generative imaging, these techniques do not come with recipes for lifting them into 3D. In this paper, we address this shortcoming by proposing a Latent Consistency Model (LCM) adaptation for the task at hand. We analyze the strengths and weaknesses of the proposed model and evaluate it quantitatively and qualitatively. Based on the Objaverse dataset samples study, our 3D painting method attains strong preference in all evaluations. Source code is available at https://github.com/kongdai123/consistency2.