Learning of Object Concept Through Function and Shape
Yosuke Sato, Takayuki Nagai · 2006
This paper discusses a novel framework for object understanding. Many conventional object learning and recognition frameworks rely upon visual information. We model object concept through the relationship between shape and function. Implementation of the proposed framework using Bayesian network is also presented. The system can form object concept by observing the human tool use. Furthermore, it is demonstrated that the learned model (object concept) enables to infer the property of unseen object. The system is evaluated using 30 hand tools, which reveals validity of the proposed framework