Stanford Seminar - Why does where people look matter? Applications of visual attention modeling

Zoya Bylinskii Adobe Research January 14, 2022 Knowing where people look has attracted the attention of many interdisciplinary scientists, with applications in computer vision, visualization, user interface and graphic design. Individual eye fixation patterns can hold clues to the intention and state of mind of the user; while pooling the eye movements of a group of individuals can reveal patterns in images and videos to help prioritize visual content for image processing and graphic design applications. I will motivate some of the automatic applications possible when computational models of visual attention are available. These include image and video cropping, as well as graphic design retargeting and reflow. At the heart of these applications and models are datasets of human visual attention patterns, many collected without explicit eye tracking, instead using a variety of crowdsourcing techniques. This talk will feature the data collection interfaces, models, and applications of human eye mo