Lecture Series in Intelligent Perception and Computing
题 目（TITLE）： Bayesian joint topic modelling for object and action detection
讲 座 人（SPEAKER）: Dr. Tao Xiang, Queen Mary University of London
主 持 人 (CHAIR)： Prof. Liang Wang
时 间 (TIME)：April 10, 2014(Thursday), 4:00pm
地 点 (VENUE)： Meeting Room, 16 Floor, Intelligent building
In this talk, I will focus on the problem of localisation of objects and actions as bounding boxes in images and videos with weak labels. This weakly supervised localisation problem has been tackled in the past using discriminative models where each class is localised independently from other classes. I will introduce a new framework we developed recently based on Bayesian joint topic modelling. The framework differs significantly from the existing ones in that all classes classes are modelled jointly in a single generative model that encodes multiple object co-existence so that ``explaining away'' inference can resolve ambiguity and lead to better learning and localisation. Moreover, the Bayesian formulation enables the exploitation of various types of prior knowledge to compensate for the limited supervision offered by weakly labelled data. With this framework, the potentially unlimited weakly labelled visual data on the Web can be exploited for various vision problems.
Tao Xiang received the PhD degree in electrical and computer engineering from the National University of Singapore in 2002. He is currently a senior lecturer (associate professor) in the School of Electronic Engineering and Computer Science, Queen Mary University of London. His research interests include computer vision, statistical learning, video processing, and machine learning, with a focus on interpreting and understanding human behavior. He has published more than 100 papers in international journals and conferences and coauthored a book, Visual Analysis of Behaviour: From Pixels to Semantics.