Pi-GS: Sparse-View Gaussian Splatting with Dense {\pi}^3 Initialization

TL;DR

This paper introduces Pi-GS, a robust method for sparse-view neural rendering that combines dense point cloud initialization with regularization techniques, enabling high-quality novel view synthesis without relying on accurate camera poses.

Contribution

It presents a novel, reference-free point cloud estimation network, Pi^3, integrated with regularization schemes to improve 3D Gaussian Splatting in sparse-view scenarios.

Findings

Achieves state-of-the-art results on multiple datasets.

Effectively handles pose and depth inaccuracies.

Enables high-quality rendering with sparse views.

Abstract

Novel view synthesis has evolved rapidly, advancing from Neural Radiance Fields to 3D Gaussian Splatting (3DGS), which offers real-time rendering and rapid training without compromising visual fidelity. However, 3DGS relies heavily on accurate camera poses and high-quality point cloud initialization, which are difficult to obtain in sparse-view scenarios. While traditional Structure from Motion (SfM) pipelines often fail in these settings, existing learning-based point estimation alternatives typically require reliable reference views and remain sensitive to pose or depth errors. In this work, we propose a robust method utilizing {\pi}^3, a reference-free point cloud estimation network. We integrate dense initialization from {\pi}^3 with a regularization scheme designed to mitigate geometric inaccuracies. Specifically, we employ uncertainty-guided depth supervision, normal consistency loss, and depth warping. Experimental results demonstrate that our approach achieves state-of-the-art performance on the Tanks and Temples, LLFF, DTU, and MipNeRF360 datasets.