Advanced Lecture Series in Pattern Recognition
题 目 (TITLE)：Beyond What and Where: Reasoning Function, Physics, Intents and Causality
讲 座 人 (SPEAKER)：Prof. Songchun Zhu (University of California, Los Angeles)
主 持 人 (CHAIR)：Prof. Liang Wang
时 间 (TIME)：July 11(Friday), 2014, 10:00AM
地 点 (VENUE)：No.1 Conference Room (3rd floor), Intelligence Building
The goal of computer vision, coined by Marr’s book published in 1982, is to compute “what” are “where” by ‘looking”. This definition has guided research on geometric approach in the 1980s-1990s and appearance approach in the 1990s -2010s. In this talk, I will argue that both geometric and appearance approaches are fundamentally limited, and vision must be integrated with AI and robotics. The tasks of object recognition, scene understanding, and event analysis should be solved by joint modeling and inference of the “visible” entities, such as geometry and appearance, and the “dark” entities: such as functionality, physics, intents, and causality. I will show a number of examples, and then propose a spatial, temporal, and causal And-or Graph (STC-AoG) as a unifying representation. I will demonstrate some recent projects on joint STC inference where the system can answer human queries of who, what, where, when, and why in a restricted Turing test.
Song-Chun Zhu graduated from USTC in 1991 and
received a Ph.D. degree from Harvard University in 1996. He is currently a professor of Statistics and Computer Science, and director of the Center for Vision, Learning, Cognition and Arts, at University of California, Los Angeles. His work in computer vision received a number of honors, including the Marr Prize in 2003 for image parsing with Z. Tu et al., the Marr Prize honorary nominations in 1999 for texture modeling and 2007 for object modeling with Y. Wu et al. As a junior faculty he received in 2001 the Sloan Fellow in Computer Science, NSF Career Award, and ONR Young Investigator Award. In 2008 he received the Aggarwal prize from the Intl Association of Pattern Recognition for “contributions to a unified foundation for visual pattern conceptualization, modeling, learning, and inference”. In 2013 he received the Helmholtz Test-of-time prize at ICCV. He is a fellow of IEEE Computer Society.