A Bayesian approach to visual size classification of everyday objects

Humans are adept at size classification from visual images of objects. A challenging computer vision problem is that of automatic visual size classification. Current size classification systems assume controlled environments and use features geared towards a particular object category and pose. However, certain applications may require algorithms that can adapt to a variety of object categories and handle complex environments. In this paper, we propose a Bayesian approach to automatic visual size classification, inspired by human visual perception, for a more generalized and robust size classifier. Initial results show that the proposed approach can handle multiple object categories and is invariant to scale changes.