Towards 3D modeling using sketches and retrieval

Retrieving 2D and 3D drawings by content is not an easy task. Automatic feature extraction, indexing and matching are some of the problems raised by these approaches. We have developed a generic method to classify, index and retrieve drawings using sketches, based on spatial relationships, shape geometry and high-dimensional indexing mechanisms. This approach has been applied with success to retrieving clip-art and complex technical drawings from large databases. In this paper we give a brief overview of our approach for content-based retrieval and describe two prototypes for retrieving 2D drawings. We also present a preliminary study that combines retrieval of 3D objects and expectation lists to define a new interaction paradigm based on suggestions. 3D objects are described using their face and edge graphs, which are then mapped into multidimensional descriptors through graph spectra. Preliminary results show that the combination of these two descriptors (faces and edges) provide a good novel method to describe and retrieve similar 3D objects. Finally, rather than developing a system to specify and display 3D queries and results, we integrated the retrieval system into a 3D modeling tool, through the use of expectation lists. This way, results from the query are presented as suggestions to the user, in what constitutes a new interaction paradigm, which is more flexible then present approaches.