Neural Surface Refinement for Modeling Transparent Objects
Weijian Deng1,
Dylan Campbell1,
Chunyi Sun1,

Shubham Kanitkar2,
Matthew E. Shaffer2,
Stephen Gould1,
1Australian National University, 2RIOS Intelligent Machines

We present Transparent Neural Surface Refinement (TNSR), which reconstructs transparent surfaces using only color consistency from multi-view RGB images, significantly improving geometry estimation and view synthesis.


Neural implicit surface reconstruction leveraging volume rendering has led to significant advances in multi-view reconstruction. However, results for transparent objects can be very poor, primarily because the rendering function fails to account for the intricate light transport induced by refraction and reflection. In this study, we introduce transparent neural surface refinement (TNSR), a novel surface reconstruction framework that explicitly incorporates physical refraction and reflection tracing. Beginning with an initial, approximate surface, our method employs sphere tracing combined with Snell's law to cast both reflected and refracted rays. Central to our proposal is an innovative differentiable technique devised to allow signals from the photometric evidence to propagate back to the surface model by considering how the surface bends and reflects light rays. This allows us to connect surface refinement with volume rendering, enabling end-to-end optimization solely on multi-view RGB images. In our experiments, TNSR demonstrates significant improvements in novel view synthesis and geometry estimation of transparent objects, without prior knowledge of the refractive index.




Ray tracing is an optimization problem

- Analytic Derivative: Proposition 4.6 from Gould et al. (2021)

Gould etal., "Deep declarative networks." In TPAMI, 2021


Enhanced Novel View Synthesis

Improved Surface Reconstruction

Paper and Supplementary Material

Deng W, Campbell D, Sun C, Kanitkar S, Shaffer M, Gould S.
Neural Surface Refinement for Modeling Transparent Objects.
In CVPR, 2024.
(hosted on [Paper])



This template was originally made by Phillip Isola and Richard Zhang for a colorful ECCV project; the code can be found here.