Abstract We propose a method to expand the visual coverage of training sets that consist of a small number of labeled examples using learned attributes. Our optimization formulation discovers category specifific attributes as well as the images that have high confifidence in terms of the attributes. In addition, we propose a method to stably capture example-specifific attributes for a small sized training set. Our method adds images to a category from a large unlabeled image pool, and leads to signifificant improvement in category recognition accuracy evaluated on a large-scale dataset, ImageNet.