FastGS: Training 3D Gaussian Splatting in 100 Seconds

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ren, Shiwei, Wen, Tianci, Fang, Yongchun, Lu, Biao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914182855655424
author Ren, Shiwei
Wen, Tianci
Fang, Yongchun
Lu, Biao
author_facet Ren, Shiwei
Wen, Tianci
Fang, Yongchun
Lu, Biao
contents The dominant 3D Gaussian splatting (3DGS) acceleration methods fail to properly regulate the number of Gaussians during training, causing redundant computational time overhead. In this paper, we propose FastGS, a novel, simple, and general acceleration framework that fully considers the importance of each Gaussian based on multi-view consistency, efficiently solving the trade-off between training time and rendering quality. We innovatively design a densification and pruning strategy based on multi-view consistency, dispensing with the budgeting mechanism. Extensive experiments on Mip-NeRF 360, Tanks & Temples, and Deep Blending datasets demonstrate that our method significantly outperforms the state-of-the-art methods in training speed, achieving a 3.32$\times$ training acceleration and comparable rendering quality compared with DashGaussian on the Mip-NeRF 360 dataset and a 15.45$\times$ acceleration compared with vanilla 3DGS on the Deep Blending dataset. We demonstrate that FastGS exhibits strong generality, delivering 2-7$\times$ training acceleration across various tasks, including dynamic scene reconstruction, surface reconstruction, sparse-view reconstruction, large-scale reconstruction, and simultaneous localization and mapping. The project page is available at https://fastgs.github.io/
format Preprint
id arxiv_https___arxiv_org_abs_2511_04283
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle FastGS: Training 3D Gaussian Splatting in 100 Seconds
Ren, Shiwei
Wen, Tianci
Fang, Yongchun
Lu, Biao
Computer Vision and Pattern Recognition
68T40(Primary)68T45, 68U99 (Secondary)
I.4.8; I.3.7
The dominant 3D Gaussian splatting (3DGS) acceleration methods fail to properly regulate the number of Gaussians during training, causing redundant computational time overhead. In this paper, we propose FastGS, a novel, simple, and general acceleration framework that fully considers the importance of each Gaussian based on multi-view consistency, efficiently solving the trade-off between training time and rendering quality. We innovatively design a densification and pruning strategy based on multi-view consistency, dispensing with the budgeting mechanism. Extensive experiments on Mip-NeRF 360, Tanks & Temples, and Deep Blending datasets demonstrate that our method significantly outperforms the state-of-the-art methods in training speed, achieving a 3.32$\times$ training acceleration and comparable rendering quality compared with DashGaussian on the Mip-NeRF 360 dataset and a 15.45$\times$ acceleration compared with vanilla 3DGS on the Deep Blending dataset. We demonstrate that FastGS exhibits strong generality, delivering 2-7$\times$ training acceleration across various tasks, including dynamic scene reconstruction, surface reconstruction, sparse-view reconstruction, large-scale reconstruction, and simultaneous localization and mapping. The project page is available at https://fastgs.github.io/
title FastGS: Training 3D Gaussian Splatting in 100 Seconds
topic Computer Vision and Pattern Recognition
68T40(Primary)68T45, 68U99 (Secondary)
I.4.8; I.3.7
url https://arxiv.org/abs/2511.04283