Regions of Interest Detection by Machine Learning Approaches

Speaker: Jun Zhou

Affiliation: Australian National University

Time: Thursday 15/09/2011 from 11:00 to 12:00

Venue: Access Grid UWS. Presented from Penrith (Y239), accessible from Parramatta (EB.1.32) and Campbelltown (26.1.50).

Abstract: Regions of interest detection (ROI) is an important step towards accurate object and image classification. In this talk, I am going to introduce our recent work to tackle the problem using machine learning approaches. I will present two models. The first model explores saliency learning approach to discover an ROI as the most salient object in an image. It divides the feature space into sub-regions of linearly separable data-points, and then uses a mixture of classifiers to recover the ROI. The second model adopts multiple instance learning approach, which deals with the classification of collections of instances called bags. Each bag, in this case as an image, contains a number of instances that represent ROI candidates extracted from local regions. If a bag has been classified as an image that contains certain class of object, then the true ROI in it can be found via instance comparison to the trained object models.

Biography: Dr Jun Zhou joined the Research School of Computer Science, the Australian National University as a research fellow in August 2011. Prior to this, he was a researcher at the Canberra Research Laboratory of NICTA (National ICT Australia Ltd). He had hold adjunct positions at the Research School of Information Sciences and Engineering in the Australian National University, and at the School of Engineering and Information Technology in the University of New South Wales at the Australian Defence Force Academy. His research interests are in statistical pattern recognition, computer vision, and machine learning with human-in-the-loop. His most recent work has focused on Bayesian inference for environmental monitoring and computer vision applications, and its implementation on high performance computing platform. Dr Zhou received a B.S. degree in computer science and a B.E. degree in international business from Nanjing University of Science and Technology, China, in 1996 and 1998, respectively. He received an M.S. degree in computer science from Concordia University, Canada, in 2002, and a Ph.D. degree from the University of Alberta, Canada, in 2006.