A Bayesian Approach to Unsupervised One-Shot Learning of Object Categories

Learning visual models of object categories notoriously requiresthousands of training examples; this is due to thediversity and richness of object appearance which requiresmodels containing hundreds of parameters. We present amethod for learning object categories from just a few images(1 ~ 5).It is based on incorporating "generic"knowledge which may be obtained from previously learntmodels of unrelated categories. We operate in a variationalBayesian framework: object categories are represented byprobabilistic models, and "prior" knowledge is representedas a probability density function on the parameters of thesemodels. The "posterior" model for an object category is obtainedby updating the prior in the light of one or more observations.Our ideas are demonstrated on four diverse categories(human faces, airplanes, motorcycles, spotted cats).Initially three categories are learnt from hundreds of trainingexamples, and a "prior" is estimated from these. Thenthe model of the fourth category is learnt from 1 to 5 trainingexamples, and is used for detecting new exemplars a setof test images.