GLINT: Modeling Scene-Scale Transparency via Gaussian Radiance Transport

TL;DR

GLINT introduces a novel Gaussian-based framework that explicitly models scene-scale transparency by separating reflected and transmitted radiance, improving the reconstruction of transparent scenes.

Contribution

It presents a new method for modeling transparency in 3D scenes by decomposing Gaussian representations and leveraging priors from a pre-trained relighting model.

Findings

GLINT outperforms prior methods in reconstructing complex transparent scenes.

The framework effectively separates reflected and transmitted radiance.

Transparency localization is bootstrapped from geometry-separation cues.

Abstract

While 3D Gaussian splatting has emerged as a powerful paradigm, it fundamentally fails to model transparency such as glass panels. The core challenge lies in decoupling the intertwined radiance contributions from transparent interfaces and the transmitted geometry observed through the glass. We present GLINT, a framework that models scene-scale transparency through explicit decomposed Gaussian representation. GLINT reconstructs the primary interface and models reflected and transmitted radiance separately, enabling consistent radiance transport. During optimization, GLINT bootstraps transparency localization from geometry-separation cues induced by the decomposition, together with geometry and material priors from a pre-trained video relighting model. Extensive experiments demonstrate consistent improvements over prior methods for reconstructing complex transparent scenes.