Data and Models for Learning Attention
ABSTRACT: Humans have remarkable capability in processing the proliferate sensory data in an accurate, reliable, and rapid manner. We developed selective attention that facilitates learning and survival by focusing limited perceptual and cognitive resources on the most relevant part of sensory input. In this talk, I will share our recent innovations on data and models, aiming at understanding and predicting visual attention.
I will first introduce our new methods to characterize complex scenes with rich semantics, and to enable large-scale collection of attention data. I will then present our deep learning model that effectively bridges the “semantic gap” in predicting where people look at. The model highlights semantic objects without any pre-trained detector, achieving a big leap towards human performance. I will also discuss how attention is affected and affects other high-level visual and emotional factors. Live demos will be shown to illustrate our findings and results.
Bio: Catherine Qi Zhao is an assistant professor in the Department of Computer Science and Engineering at the University of Minnesota, Twin Cities. Her main research interests include computational vision, machine learning, cognitive neuroscience, and mental disorders. She obtained her Ph.D. from the University of California, Santa Cruz, followed by postdoctoral training at the California Institute of Technology. Prior to joining the University of Minnesota, she was an assistant professor in the Department of Electrical and Computer Engineering and the Department of Ophthalmology at the National University of Singapore, and the principal investigator at the Visual Information Processing Lab.