3D Reconstruction of non-visible surfaces of objects from a Single Depth View -- Comparative Study

TL;DR

This paper compares two methods for reconstructing non-visible object surfaces from a single depth view, highlighting their differences in speed and accuracy using ShapeNet data.

Contribution

It provides a comparative analysis of DeepSDF and MirrorNet for single-view 3D reconstruction of occluded object surfaces.

Findings

MirrorNet is faster than DeepSDF.

MirrorNet achieves smaller reconstruction errors in most categories.

Both methods are evaluated on ShapeNet objects.

Abstract

Scene and object reconstruction is an important problem in robotics, in particular in planning collision-free trajectories or in object manipulation. This paper compares two strategies for the reconstruction of nonvisible parts of the object surface from a single RGB-D camera view. The first method, named DeepSDF predicts the Signed Distance Transform to the object surface for a given point in 3D space. The second method, named MirrorNet reconstructs the occluded objects' parts by generating images from the other side of the observed object. Experiments performed with objects from the ShapeNet dataset, show that the view-dependent MirrorNet is faster and has smaller reconstruction errors in most categories.