Towards a domain-independent function-based recognition system

We present a system which takes as input a 3-D model and produces as output a labeling indicating those portions of the shape which serve the functional requirements of some category. Presently, the reasoning is from shape using concepts of physics and causation (e.g. to infer stability). The designed form follows function representation is powerful for a number of reasons. First, the requisite knowledge of the world is organized hierarchically by a series of specifications on the shape. The interpretation of this organization for a given shape facilitates the development and evaluation of alternative hypotheses to arrive at some categorical decision. Finally, the breadth and applicability of the approach is demonstrated by the system's extendability to multiple domains where it has continued to function efficiently with very little additional knowledge.Conference ProceedingsPost-prin