Semi-supervised human-robot interactive image recognition algorithm

Abstract

Image semantics recognition is a long-standing research topic and has been used to many application areas, including medical diagnose, public security, etc. However, how to teach a social robot to have the intelligence to recognize images through user interactions still remains open and ambitious. In this paper we propose a novel framework of semi-supervised human-robot interactive image recognition. In our framework, the user first presents unlabeled images to a humanoid robot for recognition; then the robot answers the user what the image is based on a semi-supervised learning algorithm; thirdly if the robot's answer is wrong, the user correct the robot with the right label. With the learning process going on, the robot is trained to recognize more and more images with different semantic labels. The ability of "learning image semantics" makes the user feel that the robot is more like an "intelligent life". Extensive experiments and comparisons have proved the efficiency of our framework with encouraging results.