Date: 1-2pm Monday 5 November 2007
Location: 78-420 UQ School of ITEE, St Lucia Campus
Speaker: Mr Andre Guilherme Ribeiro Balan, University of Sao Paulo, Brazil
Title: HEAD (Human Encephalon Automatic Delimiter)
Abstract:
In this paper we present HEAD, the Human Encephalon Automatic Delimiter, a new and efficient method for skull-stripping in T1-weighted MRI that combines an unique histogram analysis with binary mathematical morphology. In our experiments we use real images with highly variable noise ratios and intensity non-uniformity. We evaluate our results based on manually generated true masks and the well known Jaccard metric, achieving accuracy close to 99%. We compare our method with the popular Brain Extractor Surface algorithm (BSE), which in the same experiments achieved less than 95% of accuracy.
Biography:
André Balan received his BSc degree in Computer Science from "Universidade Estadual Paulista" at Rio Claro, Brazil, and his MSc degree in Computer Science from University of Sao Paulo at Sao Carlos, Brazil. He is currently a PhD student at University of Sao Paulo, Sao Carlos, and his research interests include, mainly, image segmentation, content-based image retrival and machine learning.
***************
Date: 02.00PM, Thu 04 Oct 2007
Location: 78-420 UQ School of ITEE, St Lucia Campus
Speaker: A/Prof Anind Dey, Carnegie Mellon University, USA
Title: Challenges in Applying Context-Aware Systems to Health and Wellness
Abstract:
Context-aware systems have reached a certain level of maturity: researchers can build reliable infrastructure to support applications, and have compelling applications to build. However, in order to make context-aware systems usable by end-users, particularly for the domain of health and wellness, a number of additional steps have to be taken, including advanced modeling, improved intelligibility, and enhanced control. In this talk, I will discuss the state-of-the-art in context-aware systems, and describe how this current state is not yet sufficient for supporting usable and adoptable context-aware systems for end-users, using a number of case studies (cognitive support and diagnosis of elder drivers, cognitive support for Alzheimer's patients, and motivational support for those trying to lose weight) to illustrate these points.
Biography:
Dr. Anind K. Dey is an Assistant Professor in the Human-Computer Interaction Institute at Carnegie Mellon University. His research interests lie at the intersection of human-computer interaction and ubiquitous computing. Specifically he performs research in context-aware computing and sensor-based interactions. He has conducted research in building context-aware infrastructures (including the Context Toolkit), models for context-aware systems, applications of context-aware applications and definitions for context-aware systems. Recently his interests have focused on how to make context-aware systems usable and adoptable by end-users. Anind received a Bachelors of Applied Science in Computer Engineering from Simon Fraser University in Burnaby, Canada in 1993, and Masters of Science in AeroSpace Engineering from Georgia Tech in 1995. He received a 2nd Masters of Science and a Ph.D. in Computer Science at Georgia Tech in 2000. He was a Senior Researcher at Intel Research Berkeley from 2001-2004 and an Adjunct Assistant Professor in the EECS Department at UC Berkeley from 2002 to 2005.
***************
Date: 1-2pm Wednesday 8 August 2007
Location: 78-622 UQ School of ITEE, St Lucia Campus
Speaker: Dr Jialie Shen, Singapore Management University
Title: Efficient benchmarking of content-based image retrieval via resampling
Abstract:
While content-based image retrieval (CBIR) is a fast expanding field, and new approaches to ever more effective retrieval are frequently proposed, relatively little attention has so far been paid to the process of evaluating the effectiveness of CBIR methods. Most of the reported evaluations use standard IR evaluation methodologies, with little consideration of their statistical significance or appropriateness for CBIR, which makes it difficult to assess the precise impact of individual methods. In this talk, we present a new approach for evaluating CBIR systems which provides both efficient and statistically-sound performance evaluation. The approach is based on stratified sampling, and provides a significant improvement over existing evaluation approaches. Comprehensive experiments using our approach to evaluate a range of CBIR methods have shown that the approach reduces not only the estimation error, but also reduces the size of the test data set required to achieve specific estimation error levels significantly.
Biography:
Dr Shen is a faculty member of the School of Information Systems at Singapore Management University, Singapore. He received his PhD from the University of New South Wales (UNSW), Australia in the area of multimedia information retrieval and database access methods. He worked as a lecturer at Sydney and researcher at Glasgow for a few years, before moving to SMU. His main research interests include information retrieval, multimedia systems, database management systems, data mining, statistical machine learning and its applications.
***************
Date: 2-3pm Tuesday 22 May 2007
Location: 78-420 UQ School of ITEE, St Lucia Campus
Speaker: Prof Krithi Ramamritham, IIT Bombay, India
Title: Tracking Dynamic Boundaries: Sensors to the Rescue
Abstract:
Large scale sensor networks are being deployed for realtime monitoring applications, such as detecting leakage of hazardous material, tracking forest fires or environmental monitoring. Consider a forest fire monitoring application that involves knowing the exact region affected by the fire. Continuous update regarding the spread of the fire, its direction distance from habitats is required to expedite preventative measures. As a fire divides the forest field into two regions, one that is affected by the fire and the one that is not, these regions can be considered to be delineated by a boundary. Sensor networks are an apt solution to address the problem of tracking dynamic boundaries. Strategically deployed sensors can operate unattended (minimizing risk to human life due to the fire) and provide continuous monitoring for rapid detection and boundary estimation of fires.
In this talk, we examine the problem of tracking dynamic boundaries occurring in natural phenomena using sensor networks. Two main challenges of this problem are energy-efficient estimations from noisy observations and continuous tracking of the boundary. We propose a novel appoach which uses discrete estimations of points on the boundary based on a regression model and a smoothed-interpolation scheme to estimate the entire boundary with high confidence. Remotely placed sensor nodes produce noisy measurements of various points on the boundary using range-sensing technique. A Kalman Filter is augmented to the basic boundary estimation approach to selectively refresh the estimated boundary at a point only if it is predicted to move out of the previous estimated intervals at that point. The combination of Kalman Filter based temporal estimation with an aperiodically updated regression-based spatial estimation, allows us to provide a low overhead solution to track dynamic boundaries. Our experimental results demonstrate that DBTR is an efficient algorithm for tracking dynamic boundary that does not require prior knowledge about the dynamics of the same. In addition, the confidence band obtained from estimates at a selected few locations provides loss of coverage less than 2%.
Short Bio:
Krithi Ramamritham received the Ph.D. in Computer Science from the University of Utah. Currently he is a Chair Professor in Computer Science and Engineering and Dean of Research and Development at IIT Bombay. He is a leading researcher in timeliness and consistency issues in computer systems, in particular, databases, real-time systems, and distributed applications. His recent work addresses these issues in the context of Dynamic Data in sensor networks, embedded systems, mobile environments, and the Web. During the last few years, he has also devoted his time in the use of Information and Communication Technologies for creating tools aimed at socio-economic development. He is a board member of VLDB Foundation and ACM SIGMOD. He is on the editorial board of Real-Time Systems (Editor-in-Chief), IEEE Transactions on Mobile Computing, Distributed and Parallel Databases, World Wide Web: Internet and Web Information Systems, IEEE Transactions on Knowledge and Data Engineering, IEEE Internet Computing, IEEE Transactions on Parallel and Distributed Systems (1997 - 2000), and The VLDB Journal. He is a Fellow of the Indian National Academy of Engineering, Fellow ACM, and Fellow IEEE.
***************
Date: 1-2PM, Wednesday, 14 March
Location: 78-420 St Lucia Campus
Speaker: Dr Dawei Song, Open University, UK
Title: Introduction to the Multimedia and Information Systems (MMIS) group at the Open University
Abstract:
In this talk, I will give a brief introduction to the Knowledge Media Institute and its newly created Multimedia and Information Systems (MMIS) group. In addition to our research directions and projects, I'll also talk about how we grow up in 1 year time to become one of the major multimedia and information retrieval groups in the UK and Europe.
Biography:
Dr Dawei Song joined KMi as a senior lecturer in September, 2005. Prior to this appointment, he worked as a senior research scientist at the Distributed Systems Technology Centre (DSTC), The University of Queensland, Australia. His research interests include various theoretical and practical aspects of information retrieval (such as logical models, language modelling, high-dimensional indexing and dimensionality reduction, pseudo- and implicit relevance feedback, query expansion), text mining, applied logic, and their applications in knowledge discovery and management. In addition to textual information retrieval, he has been heavily involved in several projects/bids on multi-media (including 3D protein structures) content-based retrieval.
