We summarised the recent dataset for transparent object reconstruction(Datasets), and the state-of-the-art segmentation methods (Methods). You can directly access both the papers and their datasets via the super link.
Transparent object reconstruction methods that either reconstruct the noisy depth obtained with RGB-D cameras or reconstruct the 3D shapes of transparent objects can significantly mitigate the geometry gap between transparent objects and opaque objects, and further promotes the outreach of robot application scenarios. Transparent object reconstruction approaches can be grouped into two categories: (1) single-view approaches; (2) multiple-view approaches.
It should be noted that we will only review the reconstruction methods that can be applied in robotic scenarios. Those approaches that assume a special capturing system or known background pattern will not be discussed in this paper.