pipeline

GigaGS: Scaling up Planar-Based 3D Gaussians for Large Scene Surface Reconstruction


Junyi Chen*1,2, Weicai Ye*†1,3, Yifan Wang1,2, Danpeng Chen1,3, Di Huang1,
Wanli Ouyang1, Guofeng Zhang3, Yu Qiao1, Tong He†1

1Shanghai AI Lab, 2Shanghai Jiao Tong University, 3Zhejiang University,
*Equal Contributions †Corresponding Authors
Paper Code

Abstract

3D Gaussian Splatting (3DGS) has shown promising performance in novel view synthesis. Previous methods adapt it to obtaining surfaces of either individual 3D objects or within limited scenes. In this paper, we make the first attempt to tackle the challenging task of large-scale scene surface reconstruction. This task is particularly difficult due to the high GPU memory consumption, different levels of details for geometric representation, and noticeable inconsistencies in appearance. To this end, we propose GigaGS, the first work for high-quality surface reconstruction for large-scale scenes using 3DGS. GigaGS first applies a partitioning strategy based on the mutual visibility of spatial regions, which effectively grouping cameras for parallel processing. To enhance the quality of the surface, we also propose novel multi-view photometric and geometric consistency constraints based on Level-of-Detail representation. In doing so, our method can reconstruct detailed surface structures. Comprehensive experiments are conducted on various datasets. The consistent improvement demonstrates the superiority of GigaGS.

pipeline

Fig. 1: Overview of GigaGS. We propose GigaGS, the first work specifically designed for large scene surface reconstruction. Our approach ensures high rendering quality while also extracting high-quality meshes.

Comparison

Here, we present the qualitative comparison results of Ground Truth, GigaGS (Ours), SuGaR and Neuralangelo.

GT Ours SuGaR Neuralangelo

Ours SuGaR Neuralangelo

BibTeX

@misc{chen2024gigagsscalingplanarbased3d,
      title={GigaGS: Scaling up Planar-Based 3D Gaussians for Large Scene Surface Reconstruction}, 
      author={Junyi Chen and Weicai Ye and Yifan Wang and Danpeng Chen and Di Huang and Wanli Ouyang and Guofeng Zhang and Yu Qiao and Tong He},
      year={2024},
      eprint={2409.06685},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2409.06685}, 
}