AdpSplit: Error-Driven Adaptive Splitting for Faster Geometry Discovery in 3D Gaussian Splatting

2026-05-07

AI Summary

This work addresses the inefficiency of traditional 3D Gaussian Splatting (3DGS), which relies on a fixed number of randomly split Gaussians, leading to suboptimal geometric detail recovery and requiring multiple densification rounds that bottleneck training speed. To overcome this limitation, the authors propose a pixel error–driven adaptive splitting operator that, for the first time, incorporates L1 pixel error statistics into the Gaussian splitting process. This approach dynamically determines the number of child Gaussians and optimizes their initialization parameters based on local reconstruction error. The method substantially reduces the number of densification iterations, achieving training speedups of 9.2%–22.3% across multiple datasets. When combined with FastGS, it yields a 16.4% acceleration on MipNeRF360 while maintaining PSNR comparable to full training—resulting in a 12.6× speedup over the original 3DGS.

Abstract

Adaptive density control in 3D Gaussian Splatting (3DGS) repeatedly grows the Gaussian population through fixed-cardinality random splitting to discover useful scene structure. However, in vanilla 3DGS, its binary split operator requires many densification rounds to expose fine details, making it a bottleneck for efficient training schedules with fewer iterations. We introduce AdpSplit, an error-driven adaptive split operator that determines the number of split children and initializes the child parameters from L1-pixel-error region statistics, enabling fewer densification iterations, thus reduced training time, while preserving the rendering quality of full-schedule training. Across the MipNeRF360, Deep-Blending, and Tanks&Temples datasets, AdpSplit reduces the training time of multiple accelerated 3DGS pipelines by 9.2%-22.3% as a simple drop-in replacement for the standard split operator. With FastGS, AdpSplit matches the full-schedule PSNR on MipNeRF360 while reducing training time by 16.4%, corresponding to a 12.6x acceleration over vanilla 3DGS.

Problem

• 3D Gaussian Splatting
• adaptive splitting
• geometry discovery
• training efficiency
• densification

Innovation

• Adaptive Splitting
• Error-Driven Densification
• 3D Gaussian Splatting
• Training Acceleration
• Geometry Discovery