Learning from the Web: Recognition method based on object appearance from Internet images

Abstract

In this work an Object Learning and Recognition method for a Humanoid is presented. This method aims to take advantage of the Cloud Resources, since it is based on image web search in order to build training sets for learning objects' appearance. In case Internet access is unavailable, the robot asks human to show the objects and acquires the images using its camera. Through this technique, our method aims to be a flexible and natural framework for Human-Robot Interaction and to provide as much autonomy as possible to the robot.