Talk on Learning from Visual Data in the Wild

Title 

Abstract

The wealth and complexity of visual information potentially observable by artificial systems deployed in the real world vastly exceeds the comparative simplicity of many of our existing carefully curated computer vision benchmark datasets. To operate safely and reliably in challenging environments, artificial systems need to be able to recognize existing fine-grained visual concepts, discover new ones, and to understand the 3D structure of the world around them. In this talk I will discuss some recent work from my group on problems related to learning from visual data “in the wild”. Specifically, I will cover recent work on the self-supervised learning of 3D shape from images, category discovery, and how to encode geographic priors about where objects are likely to be observed.  

 

Bio

Oisin Mac Aodha is a Lecturer in Machine Learning in the School of Informatics at the University of Edinburgh. He is also a Turing Fellow and ELLIS Scholar. He obtained his PhD from University College London, advised by Gabriel Brostow, and was a post-doc with Pietro Perona at Caltech before joining the University of Edinburgh. His current research interests are in the areas of self-supervised learning, 3D vision, fine-grained learning, and human-in-the-loop learning. More information can be found on his website: https://homepages.inf.ed.ac.uk/omacaod.