DashGaussian: Optimizing 3D Gaussian Splatting in 200 Seconds

2025-03-24

AI Summary

To address computational redundancy and excessive training time in 3D Gaussian Splatting (3DGS) optimization—caused by high rendering resolutions and an overabundance of Gaussian primitives—this work formulates optimization as a multi-band progressive fitting task. We propose a frequency-aware dynamic scheduling mechanism that jointly controls rendering resolution and Gaussian primitive count: during low-frequency stages, coarse resolution and sparse primitives enable rapid convergence; in high-frequency stages, both resolution and density are progressively increased to refine geometric and appearance details. Our method achieves state-of-the-art rendering quality—preserving PSNR and SSIM without degradation—while accelerating optimization by 45.7% on average and reducing total training time to under 200 seconds. The approach is fully compatible with mainstream 3DGS backbone architectures, significantly enhancing training efficiency without compromising fidelity.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Computer Vision: 3D Computer Vision
• Machine Learning: Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

3D Gaussian Splatting (3DGS) renders pixels by rasterizing Gaussian primitives, where the rendering resolution and the primitive number, concluded as the optimization complexity, dominate the time cost in primitive optimization. In this paper, we propose DashGaussian, a scheduling scheme over the optimization complexity of 3DGS that strips redundant complexity to accelerate 3DGS optimization. Specifically, we formulate 3DGS optimization as progressively fitting 3DGS to higher levels of frequency components in the training views, and propose a dynamic rendering resolution scheme that largely reduces the optimization complexity based on this formulation. Besides, we argue that a specific rendering resolution should cooperate with a proper primitive number for a better balance between computing redundancy and fitting quality, where we schedule the growth of the primitives to synchronize with the rendering resolution. Extensive experiments show that our method accelerates the optimization of various 3DGS backbones by 45.7% on average while preserving the rendering quality.

Problem

• Accelerating 3DGS optimization by reducing redundant complexity
• Dynamic rendering resolution for lower optimization complexity
• Synchronizing primitive growth with rendering resolution for balance

Innovation

• Dynamic rendering resolution reduces optimization complexity
• Synchronized primitive growth with rendering resolution
• Accelerates 3DGS optimization by 45.7%