PairDropGS: Paired Dropout-Induced Consistency Regularization for Sparse-View Gaussian Splatting

2026-05-12

AI Summary

Existing Dropout-based sparse-view 3D Gaussian splatting methods suffer from reconstruction instability and suboptimal representation learning due to neglecting inconsistencies among different randomly dropped subsets. To address this, this work proposes a paired Dropout mechanism that constructs two subsets sharing a common Gaussian field and introduces a low-frequency consistency regularization to stabilize scene layout and coarse geometry while preserving high-frequency details. Combined with a progressive consistency scheduling strategy, the approach significantly enhances training stability and reconstruction quality without compromising methodological simplicity. The proposed method consistently outperforms existing Dropout-based approaches across multiple sparse-view benchmarks and offers a plug-and-play capability for optimization enhancement.

Abstract

Dropout-based sparse-view 3D Gaussian Splatting (3DGS) methods alleviate overfitting by randomly suppressing Gaussian primitives during training. Existing methods mainly focus on designing increasingly sophisticated dropout strategies, while they overlook the resulting inconsistencies among different dropped Gaussian subsets. This oversight often leads to unstable reconstruction and suboptimal Gaussian representation learning.In this paper, we revisit dropout-based sparse-view 3DGS from a consistency regularization perspective and propose PairDropGS, a Paired Dropout-induced Consistency Regularization framework for sparse-view Gaussian splatting. Specifically, PairDropGS first constructs a pair of the dropped Gaussian subsets from a shared Gaussian field and designs a low-frequency consistency regularization to constrain their low-frequency rendered structures. This design encourages the shared Gaussian field to preserve stable scene layout and coarse geometry under different random dropouts, while avoiding excessive constraints on ambiguous high-frequency details. Moreover, we introduce a progressive consistency scheduling strategy to gradually strengthen the consistency regularization during training for stability and robustness of reconstruction. Extensive experiments on widely-used sparse-view benchmarks demonstrate that PairDropGS achieves superior training stability, significantly outperforms existing dropout-based 3DGS methods in reconstruction quality, while exhibiting the simplicity and plug-and-play nature for improving dropout-based optimization.

Problem

• sparse-view
• 3D Gaussian Splatting
• dropout
• consistency regularization
• overfitting

Innovation

• consistency regularization
• paired dropout
• sparse-view reconstruction
• 3D Gaussian Splatting
• low-frequency constraint