Over the past two decades, the advent of the World Wide Web and improvements in technology for the acquisition and visualization of 3D models have combined to effect an explosion in the availability of 3D models to everyday users. In this context, the challenge has changed from “How do we generate 3D models?” to “How do we find them?”
In this talk, we describe research in the area of shape matching. We provide a general overview of approaches aimed at addressing this challenge and focus specifically on the difficulty of matching 3D models across different orientations. In particular, we present a method for obtaining rotation invariant representations of 3D models. We show that this method is well suited to the task of shape matching by providing a compact representation that allows for efficient and discriminating retrieval from large repositories of 3D shapes.