FDGaussian: Fast Gaussian Splatting from Single Image via Geometric-aware Diffusion Model

Arxiv 2024

Qijun Feng, Zhen Xing, Zuxuan Wu, Yu-Gang Jiang

Fudan Univerisity  

Arxiv Code

Abstract

Reconstructing detailed 3D objects from single-view images remains a challenging task due to the limited information available. In this paper, we introduce FDGaussian, a novel two-stage framework for single-image 3D reconstruction. Recent methods typically utilize pre-trained 2D diffusion models to generate plausible novel views from the input image, yet they encounter issues with either multi-view inconsistency or lack of geometric fidelity. To overcome these challenges, we propose an orthogonal plane decomposition mechanism to extract 3D geometric features from the 2D input, enabling the generation of consistent multi-view images. Moreover, we further accelerate the state-of-the-art Gaussian Splatting incorporating epipolar attention to fuse images from different viewpoints. We demonstrate that FDGaussian generates images with high consistency across different views and reconstructs high-quality 3D objects, both qualitatively and quantitatively.

Novel View Generation and 3D Reconstruction

Qualitative Comparison

Interactive Results

panda
windmill
Image 3
Image 1
Image 2
Image 3
Image 1
Image 2
Image 3

Text-to-image-to-3D

A grey double sofa.

Image 1

A basket of bread.

Image 1

A telephone in vintage style.

Image 1

Citation 

@misc{feng2024fdgaussian,
          title={FDGaussian: Fast Gaussian Splatting from Single Image via Geometric-aware Diffusion Model}, 
          author={Qijun Feng and Zhen Xing and Zuxuan Wu and Yu-Gang Jiang},
          year={2024},
          eprint={2403.10242},
          archivePrefix={arXiv},
          primaryClass={cs.CV}}