Robot painting recognition based on deep belief learning

Abstract

In a society where the number of elderly people is increasing rapidly, autonomous wheelchair robots are expected to be widely used for mobility of elderly people. In this paper we focus on how we can utilize wheelchair robots operating in museums. In this paper, we propose a deep learning based painting recognition and its application for the wheelchair robot. We consider the case when the user clicks on the painting he/she wants to see. The robot searches, recognizes and reaches the painting using deep learning. This is in difference from the most traditional methods where the robot explains the exhibited objects in a sequential order. The deep neural network generates a series of high dimensional features for each painting resulting in a high recognition rate. In our implementation, the wheelchair robot recognizes the painting in real time using the video stream.