Interactive Viewpoint Suggestion with Deep Classification Network for Browsing 3D Object Galleries

Anonymous

Keywords: Visualization, Deep Learning, interaction

TL;DR: A new framework for simultaneously selecting viewpoints to compare multiple objects in 3D galleries.

Abstract: Viewpoint selection techniques have been one of the most fundamental and main topics in computer graphics for many years and closely related to some other applications, such as data visualization or camera placement. Also, according to the rapid improvements in the deep learning field, some viewpoint selection techniques based on deep learning methods have been proposed. However, these methods are typically only focused on suggesting one viewpoint of a single object. In this paper, we propose a new framework for simultaneously selecting viewpoints to compare multiple objects in 3D galleries. Our network takes rendered images of each object from various viewpoints as inputs, and outputs optimal viewpoints for each object so that users can easily grasp characteristics of 3D objects. Furthermore, for more general-purpose usage, the system also supports interactions with users, so that users can easily explore the viewpoints of some objects. We validated the efficiency of our approach through the user study.

Track: HCI/visualization