NeTO: Neural Reconstruction of Transparent Objects
with Self-Occlusion Aware Refraction-Tracing


We present a novel method called NeTO, for capturing the 3D geometry of solid transparent objects from 2D images via volume rendering. Reconstructing transparent objects is a very challenging task, which is ill-suited for general-purpose reconstruction techniques due to the specular light transport phenomena. Although existing refraction-tracing-based methods, designed especially for this task, achieve impressive results, they still suffer from unstable optimization and loss of fine details since the explicit surface representation they adopted is difficult to be optimized, and the self-occlusion problem is ignored for refraction-tracing. In this paper, we propose to leverage implicit Signed Distance Function (SDF) as surface representation and optimize the SDF field via volume rendering with a self-occlusion aware refractive ray tracing. The implicit representation enables our method to be capable of reconstructing high-quality reconstruction even with a limited set of views, and the self-occlusion aware strategy makes it possible for our method to accurately reconstruct the self-occluded regions. Experiments show that our method achieves faithful reconstruction results and outperforms prior works by a large margin.

Comparisons on DRT dataset with sparse views

To validate the ability of NeTO to reconstuct transparent surfaces, we conduct comparisons with current state-of-the-art method on the DRT dataset and our real data. dataset. As you can see, with sparse views, our results outperform DRT in terms of model completeness and accuracy.

Comparisons on DRT dataset with Full views

When we make use of more views, e.g., full views, the reconstruction results of ours and DRT are improved compared with reconstruc- tions with sparse views.

Comparisons with DRT on our real data

We further conduct evaluation on a self-collected real Bull and Mouse object. Our method accurately recovers the geometry with clean and smooth surfaces, while DRT mistakenly reconstructs surfaces with noises.


  title={NeTO: Neural Reconstruction of Transparent Objects with Self-Occlusion Aware Refraction-Tracing},
  author={Li, Zongcheng and Long, Xiaoxiao and Wang, Yusen and Cao, Tuo and Wang, Wenping and Luo, Fei and Xiao, Chunxia},
  journal={arXiv preprint arXiv:2303.11219},

This page is Zotero translator friendly. Page last updated