Next best view estimation for volumetric information gain

In this work we propose a novel next best view (NBV) generation algorithm for volumetric information maximization. The primary data source is a Time-of-Flight (ToF) camera and the output is the next position of the depth sensor that maximizes a chosen score, either coverage or histogram of volumetric estimations for parallelepipedic shapes on the observed scene. Our learning-based method was validated on a large scale of real and synthetic data. The demo code, custom data sets, and videos are available on the author's website.