Differentiable Surface Splatting for Point-based Geometry Processing

TL;DR

This paper introduces Differentiable Surface Splatting (DSS), a novel high-fidelity differentiable renderer for point clouds that enables accurate inverse rendering and geometry processing without explicit connectivity.

Contribution

The paper presents DSS with carefully designed gradients and regularization, improving inverse rendering and denoising of point clouds over existing methods.

Findings

Outperforms state-of-the-art techniques in inverse rendering tasks.

Handles large topological changes and small detail modifications robustly.

Provides open-source code and data for reproducibility.

Abstract

We propose Differentiable Surface Splatting (DSS), a high-fidelity differentiable renderer for point clouds. Gradients for point locations and normals are carefully designed to handle discontinuities of the rendering function. Regularization terms are introduced to ensure uniform distribution of the points on the underlying surface. We demonstrate applications of DSS to inverse rendering for geometry synthesis and denoising, where large scale topological changes, as well as small scale detail modifications, are accurately and robustly handled without requiring explicit connectivity, outperforming state-of-the-art techniques. The data and code are at https://github.com/yifita/DSS.