# Unique3D: High-Quality and Efficient 3D Mesh Generation from a Single Image

**Authors:** Kailu Wu, Fangfu Liu, Zhihan Cai, Runjie Yan, Hanyang Wang, Yating Hu, Yueqi Duan, Kaisheng Ma  
**arXiv:** [2405.20343](https://arxiv.org/abs/2405.20343)  
**Date:** 2024-05-30  
**Summary page:** https://www.summarizepaper.com/en/paper/2405.20343v2/

## Key points

- Unique3D is a groundbreaking image-to-3D framework that sets a new standard in high-fidelity, efficiency, and generalizability for generating 3D meshes from single-view images.
- The approach leverages a multi-view diffusion model and a normal diffusion model to generate four orthographic multi-view images with corresponding normal maps, progressively upscaled to higher resolutions.
- Key contribution: ISOMER, an instant and consistent mesh reconstruction algorithm integrating color and geometric priors for detailed 3D meshes with intricate textures and complex geometries.
- Superior performance demonstrated through extensive experiments compared to existing image-to-3D baselines in terms of geometric accuracy and textural details.
- Unlocking new possibilities for real-world applications in 3D generative AI by achieving a balance between high fidelity, consistency, and efficiency in single image-to-3D conversion tasks.

## Summary

, , , , 

Unique3D is a groundbreaking image-to-3D framework that sets a new standard in high-fidelity, efficiency, and generalizability for generating 3D meshes from single-view images. Our approach builds upon previous methods such as Score Distillation Sampling (SDS), which extract 3D knowledge from large 2D diffusion models but suffer from long optimization times and inconsistency issues. To overcome these limitations, we leverage a multi-view diffusion model and a normal diffusion model to generate four orthographic multi-view images with corresponding normal maps. These images are then progressively upscaled to higher resolutions. One of the key contributions of Unique3D is the development of ISOMER, an instant and consistent mesh reconstruction algorithm that integrates color and geometric priors. This allows us to reconstruct detailed 3D meshes with intricate textures and complex geometries. Through extensive experiments, we demonstrate the superior performance of Unique3D compared to existing image-to-3D baselines in terms of both geometric accuracy and textural details. Furthermore, our method unlocks new possibilities for real-world applications in the field of 3D generative AI by achieving a balance between high fidelity, consistency, and efficiency in single image-to-3D conversion tasks. By combining cutting-edge techniques such as multi-level upscaling and instant mesh reconstruction, Unique3D showcases the potential for advancing the state-of-the-art in 3D mesh generation from single images.