CityGaussianV2

CityGaussianV2: Efficient and Geometrically Accurate Reconstruction for Large-Scale Scenes

This model, presented in CityGaussianV2: Efficient and Geometrically Accurate Reconstruction for Large-Scale Scenes, offers efficient and high-fidelity 3D reconstruction of large-scale scenes. It addresses challenges in geometric accuracy and scalability by utilizing a decomposed-gradient-based densification and depth regression technique, along with an elongation filter to mitigate Gaussian count explosion. CityGaussianV2 achieves significant improvements in speed, memory usage, and visual quality compared to previous methods.

<p align="center"> <img width="460" height="300" src="assets/demo.gif"> </p>

Project page: https://dekuliutesla.github.io/CityGaussianV2/

Code: https://github.com/DekuLiuTesla/CityGaussian

Checkpoints: Checkpoints are available via Baidu Netdisk and Hugging Face (links provided in the original README).

Citation

@misc{liu2024citygaussianv2efficientgeometricallyaccurate,
      title={CityGaussianV2: Efficient and Geometrically Accurate Reconstruction for Large-Scale Scenes}, 
      author={Yang Liu and Chuanchen Luo and Zhongkai Mao and Junran Peng and Zhaoxiang Zhang},
      year={2024},
      eprint={2411.00771},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2411.00771}, 
}

@inproceedings{liu2025citygaussian,
  title={Citygaussian: Real-time high-quality large-scale scene rendering with gaussians},
  author={Liu, Yang and Luo, Chuanchen and Fan, Lue and Wang, Naiyan and Peng, Junran and Zhang, Zhaoxiang},
  booktitle={European Conference on Computer Vision},
  pages={265--282},
  year={2025},
  organization={Springer}
}