Wang, Yizhou is a professor in the Department of Computer Science and technology, School of EECS, and has served as the Vice Director of the Institute of Digital Media since 2010. He obtained his Bachelor degree from Tsinghua University in 1996, and Ph.D. from University of California at Los Angeles (UCLA) in 2005. His research interests include computer vision and statistical learning.
Dr. Wang has published over 100 research papers, and most of them are published in top-tier conferences and journals, such as IEEE-TPAMI, IJCV, CVPR, ICCV and NIPS. He is an associate editor of IEEE Trans. on Cognitive and Developmental Systems. He has served as co-program chair and member of organizing committee for a number of international prestigious academic conferences. He was granted “the National Natural Science Funds for Distinguished Young Scholar” in 2016.
Dr. Wang has more than ten research projects including NSFC, 973 programs, 863 project, etc. His research achievements are summarized as follows:
1) Attentional mechanisms and computational models: Discovering and modeling the key visual perceptual processes have both scientific and practical values. Dr. Wang and his collaborators proposed novel psychophysical and the physiological approaches to discover the neural basis of bottom-up saliency map of natural images. Based on the findings a new computational model of visual saliency was devised from the information maximization principle. He also proposed a novel computational model to simulate the dynamic trait of human attention on natural images by integrating both bottom-up and top-down perceptual mechanisms.
2) Structured sparsity pursuit: This is a key feature of natural image content. He proposed efficient methods to enforce structured sparsity prior on foreground and low rank prior on background so that figures can be accurately separated from backgrounds. The proposed optimization algorithms reveal the underlying connection between discrete optimization and continuous optimization.
3) Object and scene structure analysis: This is a core area of current computer vision research. Many convincing psychological evidences suggested that parts play a significant role in object/scene perception and recognition. He introduced perceptual cues to decompose objects into semantic parts and proposed a computational model to measure “perceptual importance” of the parts and harness the measure to improve object detection/recognition. He also proposed a weakly supervised method to learn hierarchical structures of natural scenes for image understanding. The method suggests a novel means of turning the notorious hard structure learning problem into a much easier parameter learning problem.