TL;DR We introduce Turbo-GS, which employs frequent and effective densification, better initialization, and batched training to achieve higher quality results with fewer iterations.
If you find our work helpful, please consider citing:
@misc{turbogs2024,
title={Turbo-GS: Accelerating 3D Gaussian Fitting for High-Quality Radiance Fields},
author={Tao Lu and Ankit Dhiman and R Srinath and Emre Arslan and Angela Xing and Yuanbo Xiangli and R Venkatesh Babu and Srinath Sridhar},
year={2024},
eprint={2412.13547},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2412.13547},
}