Efficient Learning of Qualitative Descriptions for Sketch Recognition

We are trying to solve the problem of learning to recognize objects in an open-domain sketching environment. Our system builds generalizations of objects based upon previous sketches of those objects and uses those generalizations to classify new sketches. We represent sketches qualitatively because we believe qualitative information provides a level of description that abstracts away details that distract from classification, such as exact dimensions. Bayesian reasoning is used in the process of building up representations to deal with the inherent uncertainty in the perception problem. Qualitative representations are compared using SME, a computational model of analogy and similarity that is supported by psychological evidence from studies of perceptual similarity. We produce generalizations based on the common structure found by SME in different sketches of the same object. We report on the results of testing the system on a corpus of sketches of everyday objects, drawn by ten