CIGMO: Categorical invariant representations in a deep generative framework

Haruo Hosoya

Abstract: Data of general object images have two most common structures: (1) each object of a given shape can be rendered in multiple different views, and (2) shapes of objects can be categorized in such a w...