Last update: Jan. 30, 2004
Learning visual models of object categories notoriously requires thousands of training examples; this is due to the diversity and richness of object appearance which requires models containing hundreds of parameters. We present a method for learning object categories from just a few images (1~5). It is based on incorporating "generic" knowledge which may be obtained from previously learnt models of unrelated categories. We operate in a variational Bayesian framework: object categories are represented by probabilistic models, and "prior" knowledge is represented as a probability density function on the parameters of these models. The "posterior" model for an object category is obtained by updating the prior in the light of one or more observations. Our ideas are demonstrated on objects belonging to 101 widely varied categories. Two learning schemes within the variational Bayesian framework are tested: batch learning and incremental learning. We show that the variational Bayes methods are capable of learning a very wide range of object categories using very few training images. While the batch and incremental methods have comparable performances, the incremental one has much shorter learning time.
10:00-11:00 ???
11:00-12:00 Taubin (B&H 328)
12:00-1:00 Talk (Savarese; B&H 190)
1:00-2:00 Lunch
2:00-3:00 Black (CS 521)
3:00-4:00 ???
4:00-5:00 Demos at LEMS (B&H 317) and SHAPE Lab. (B&H 313)
10:00-10:30 Mundy (B&H 351)
10:30-11:00 ???
11:00-12:00 ???
12:00-1:00 Talk (Li; B&H 190)
1:00-2:00 Lunch
2:00-2:45 Tarr (190 Thayer St., Rm. 230)
3:00-4:00 Cooper (B&H 318)
4:00-5:00 Kimia (B&H 320)