GLINT: Modeling Scene-Scale Transparency via Gaussian Radiance Transport
This addresses a specific limitation in 3D reconstruction for scenes with transparent objects, though it appears incremental as it builds on existing Gaussian splatting paradigms.
The paper tackles the problem of modeling scene-scale transparency in 3D Gaussian splatting, which fundamentally fails to handle transparent interfaces like glass panels, by introducing GLINT, a framework that reconstructs primary interfaces and models reflected and transmitted radiance separately, achieving consistent improvements over prior methods.
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.