Joint Seminar Series on Culture Technology & Entertainment Engineering
● Date: Feb. 21, 2008
● Place: GSCT CT Lecture room (N8, #.3229)
● Title: Real Time Computational Attention Tracking for Virtual Environments
● Speaker: Gerard Jounghyun Kim
This talk presents a real-time framework for computationally tracking visual attention, in interactive virtual environments, which uses top-down (goal-directed) context to predict human gaze from bottom-up (stimulus-driven) salient objects. The framework first builds feature maps using preattentive features such as luminance, hue, depth, size, and motion. The feature maps are integrated to the single saliency map using the center-surround difference, which is accelerated using the hardware mipmap generation in the GPU. The bottom-up saliency map is converted to an object saliency map using the item buffer. For the selected attentive candidates, the topdown context is used to select closer ones to human intention using spatial and temporal interest during interactive navigation. The framework can be used effectively, without employing an expensive eye-tracker, for perceptually based rendering such as giving depth of field effects and managing LOD.
With time permitting, other research results at Korea University's Digital Exprience Lab will be presented as well.
Gerard Jounghyun kim is an associate Professor of Dept. of Computer Science and Engineering, Korea University.
His research interests cover Human Computer Interaction, Virtual/Mixed Reality and Computer Music. He received his Ph.D in Computer Science in University from Southern California,M.S. degree in Computer Engineering from University of Southern California and B.S in Electrical & Computer Engineering from Carnegie Mellon University. He was an assistant/associate professor in Pohang University of Science and Technology(POSTECH) in 1996-2006 and a visiting scholar of Integrated Media Systems Center in University of Southern California in 2003-2004.He is a member of ACM and IEEE. He is also a program Committee Member of IEEE VR, ACM VRST, etc.