GS-2DGS: Geometrically Supervised 2DGS for Reflective Object Reconstruction

TL;DR

GS-2DGS introduces a novel 3D reflective object reconstruction method that combines 2D Gaussian Splatting with geometric constraints, achieving high-quality, fast, and detailed results surpassing previous Gaussian techniques.

Contribution

The paper presents a new geometric supervision approach for 2D Gaussian Splatting, significantly improving reflective object reconstruction speed and quality.

Findings

Outperforms Gaussian-based methods in reconstruction and relighting.

Achieves comparable quality to SDF-based methods with much faster speed.

Demonstrates effectiveness on synthetic and real datasets.

Abstract

3D modeling of highly reflective objects remains challenging due to strong view-dependent appearances. While previous SDF-based methods can recover high-quality meshes, they are often time-consuming and tend to produce over-smoothed surfaces. In contrast, 3D Gaussian Splatting (3DGS) offers the advantage of high speed and detailed real-time rendering, but extracting surfaces from the Gaussians can be noisy due to the lack of geometric constraints. To bridge the gap between these approaches, we propose a novel reconstruction method called GS-2DGS for reflective objects based on 2D Gaussian Splatting (2DGS). Our approach combines the rapid rendering capabilities of Gaussian Splatting with additional geometric information from foundation models. Experimental results on synthetic and real datasets demonstrate that our method significantly outperforms Gaussian-based techniques in terms of reconstruction and relighting and achieves performance comparable to SDF-based methods while being an order of magnitude faster. Code is available at https://github.com/hirotong/GS2DGS