Abstract: Current approaches to recognition rely on large labeled datasets. This reliance is problematic for rare classes as well as for domains where annotating data requires expertise. I will talk about my work with students and collaborators towards building new kinds of recognition systems that rely on few or no labels. I will argue that the key here is to leverage the known structure of the world behind the data points.
Bio: Bharath Hariharan is an assistant Professor of Computer Science at Cornell University. His research interests are in building versatile Computer vision systems with little training data. He is a recipient of an NSF CAREER and a PAMI Young Researcher Award.