Automatic Object Modeling by Observing Human-Object Interaction

We propose a method for constructing object appearance models by observing human-object interaction in a household environment. Using this model, the system is able to perform object recognition by feature matching. In particular, our system gradually collects different appearances in order to improve its recognition performance. We also propose an aspect selection algorithm that avoids collecting unnecessary appearances by selecting only the ones that contain relevant visual information. We performed on-line experiments with several objects and demonstrate the improvement of the recognition performance.

Demo Videos

Demo Demo Demo

Related Publications

Christian I. Penaloza, Y. Mae, K. Ohara, T. Takubo, T. Arai, "Automatic Object Modeling by Observing Human-Object Interaction", The 7th International Conference on Ubiquitous Robots and Ambient Intelligence (URAI) , pp.433-436, Pusan Korea. November 24-27, 2010.