This project proposes a Deep Learning-based approach to solving the Next Best View (NBV) planning problem. One of the fundemantal research goals of Computer Graphics is to create 3D models of objects and entire scenes. The first step of the 3D modelling pipeline is scanning. Since a single scan usually does not contain enough information about object/scene of interest and scans are susceptible to noise, multiple scans from different views are necessary. Here, NBV planning emerges as an important problem: what sequence of views is optimal?