Juan Carlos Niebles
Title- Event Understanding: a Cornerstone of Visual Intelligence
Abstract: As events are central to human life, computer vision systems of the future will need to recognize and understand human events at various levels of detail. Many of these events are naturally organized by part-of hierarchical relationships in a partonomy. We argue that leveraging such partonomystructure will be key to the success of human event understanding algorithms. In this talk, we summarize our recent progress in algorithms for recognizing and parsing events that go from simple atomic actions to complex long term tasks. At the scale of atomic actions, we will discuss an efficient algorithm for video action classification that leverages spatio-temporal shift operations instead of convolutions. In the intermediate scale, we study actions as compositions of spatio-temporal scene graphs and show how they can both improve action recognition and enable better few-shot learning of actions. When such supervised compositions are not available, we show how to leverage data-driven spatio-temporal graphs for video captioning. At the long term task scale, we discuss a method for learning the structure of tasks from instructional videos and show its applications to procedure planning. We conclude by discussing some of the open challenges that remain to fully embrace the partonomyperspective for event understanding.
Bio: Juan Carlos Nieblesreceived an Engineering degree in Electronics from Universidad del Norte (Colombia) in 2002, an M.Sc. degree in Electrical and Computer Engineering from University of Illinois at Urbana-Champaign in 2007, and a Ph.D. degree in Electrical Engineering from Princeton University in 2011. He is co-Director of the Stanford Vision and Learning Lab, Associate Director of Research at the Stanford-Toyota Center for AI Research and a Senior Research Scientist at the Stanford AI Lab since 2015. He was also an Associate Professor of Electrical and Electronic Engineering in Universidad del Norte (Colombia) between 2011 and 2019. His research interests are in computer vision and machine learning, with a focus on visual recognition and understanding of human actions and activities, objects, scenes, and events. He has served as Area Chair for the top computer vision conferences CVPR and ICCV. He is also a member of the AI Index Steering Committee and is the Curriculum Director for Stanford-AI4ALL. He is a recipient of a Google Faculty Research award (2015), the Microsoft Research Faculty Fellowship (2012), a Google Research award (2011) and a Fulbright Fellowship (2005).