Autonomous Robot Photographer System based on Aesthetic Composition Evaluation using Yohaku
Kai Lan, Kosuke Sekiyama · 2019
In this study, we aim to propose a robot photographer system which could take photographs for people with high aesthetic values by searching for the optimal photographing viewpoint autonomously. Aesthetic composition evaluation and observing viewpoint selection are two main issues which should be considered to construct this system. Different from conventional composition evaluation methods which are following some well-known rules such as rule of thirds and diagonal rule, we focus on yohaku of photographs, which means the perceptual empty space in Japanese, to help to construct an aesthetic composition more flexibly. To achieve this purpose, we first propose the yohaku detection method for actual photographs. Then aesthetic composition evaluation will be executed from the viewpoint of visual balance between ordinary targets and yohaku. Finally, using the viewpoint selection method called Composition Map, autonomous photographing will be achieved by evaluating the aesthetic value of scenes with our system.