Pope, Arthur R. and David G. Lowe, "Learning probabilistic appearance models for object recognition," in Early Visual Learning, eds. Shree Nayar and Tomaso Poggio (Oxford University Press, 1996), pp. 67--97.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....pairwise dependencies between features of the object. We then use maximum a posteriori (MAP) estimation to find the match between the object and the scene or to show that there is no such match. While a number of probabilistic approaches to recognition have been reported in the literature (e.g. [8], 7] 10] these methods do not provide an explicit model of dependencies between features. We show that finding the best match using the Hausdorff fraction [4] 9] is a special case of our technique, where features in the object model are independent. Therefore, our Bayesian framework can be ....

Arthur Pope and David G. Lowe. Learning probabilistic appearance models for object recognition. In Shree K. Nayar and Tomaso Poggio, editors, Early Visual Learning, pages 67--98. Oxford University Press, 1996.

....or object class, rather than its appearance. Another line of research, which falls midway between this approach and image based schemes, instead attempts to learn a small set of characteristic views, each of which can be used to recognize an object from a different perspective (e.g. Gros 1993; Pope Lowe 1996). Most work on visual learning ignores the importance of misclassification costs, but our work along these lines has some precedents. In particular, Draper, Brodley, and Utgoff (1994) incorporate the cost of errors into their algorithm for constructing and pruning multivariate decision trees. They ....

Pope, A., and Lowe, D. 1996. Learning probabilistic appearance models for object recognition. In Nayar, S., and Poggio, T., eds., Early visual learning. New York, NY: Oxford University Press.

....could be combined into a single model, thereby increasing the probabilityof finding matches in new views. The models could be true 3D representations based on structure from motion solutions, or could represent the space of appearance in terms of automated clustering and interpolation (Pope Lowe [17]) An advantage of the latter approach is that it could also model non rigid deformations. The recognition performance could be further improved by adding new SIFT feature types to incorporate color, texture, and edge groupings, as well as varying feature sizes and offsets. Scale invariant edge ....

Pope, Arthur R. and David G. Lowe, "Learning probabilistic appearance models for object recognition," in Early Visual Learning, eds. Shree Nayar and Tomaso Poggio (Oxford University Press, 1996), pp. 67--97.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC