GaussianPro: 3D Gaussian Splatting with Progressive Propagation
This addresses a bottleneck in neural rendering for large-scale scenes, offering a domain-specific incremental improvement over 3DGS.
The paper tackles the problem of 3D Gaussian Splatting (3DGS) struggling with large-scale scenes containing texture-less surfaces due to poor initialization from Structure-from-Motion, by proposing GaussianPro, which uses progressive propagation to guide densification, resulting in a 1.15dB PSNR improvement on the Waymo dataset.
The advent of 3D Gaussian Splatting (3DGS) has recently brought about a revolution in the field of neural rendering, facilitating high-quality renderings at real-time speed. However, 3DGS heavily depends on the initialized point cloud produced by Structure-from-Motion (SfM) techniques. When tackling with large-scale scenes that unavoidably contain texture-less surfaces, the SfM techniques always fail to produce enough points in these surfaces and cannot provide good initialization for 3DGS. As a result, 3DGS suffers from difficult optimization and low-quality renderings. In this paper, inspired by classical multi-view stereo (MVS) techniques, we propose GaussianPro, a novel method that applies a progressive propagation strategy to guide the densification of the 3D Gaussians. Compared to the simple split and clone strategies used in 3DGS, our method leverages the priors of the existing reconstructed geometries of the scene and patch matching techniques to produce new Gaussians with accurate positions and orientations. Experiments on both large-scale and small-scale scenes validate the effectiveness of our method, where our method significantly surpasses 3DGS on the Waymo dataset, exhibiting an improvement of 1.15dB in terms of PSNR.