Active Random Forests: An application to Autonomous Unfolding of Clothes
We present Active Random Forests, a novel framework to address active vision problems. State of the art focuses on best viewing parameters selection like viewpoint or zooming based on single view classifiers. In contrast, we propose a multi-view classifier where the action taking process about optimally selecting viewing parameters is inherent to the classification process. This has many advantages: a) The classifier exploits the entire set of images captured at a certain time and does not simply aggregate probabilistically per view hypotheses; b) actions are made according to learnt disambiguating image features from all possible views and are optimally selected using the powerful voting scheme of Random Forests and c) the classifier can take into account the costs of its actions. The proposed framework is applied to the task of autonomously unfolding clothes by a robot, addressing the problem of best viewpoint selection in classification, pose and grasp point estimation of garments. We show great performance improvement compared to random viewpoint selection and state of the art methods.