home · mobile · calendar · colloquia · 2000-2001 · 

Colloquium - Burl

Mining Large Image Collections
Michael Burl
Jet Propulsion Laboratory

Improvements in acquisition and storage technology have led to an explosion in the number and size of image collections in a variety of fields from medical imaging to the petroleum industry to digital libraries and the Internet to space exploration. Within these datasets there is a potential wealth of information; however, transforming from the raw data (perhaps millions of images each containing millions of pixels) to a higher-level understanding of the content of an image collection is a difficult task both due to the size of the datasets and the difficulty of automatically interpreting image data. Recent attempts to approach the problem using a distributed set of human labelers [ClickWorkers] are interesting, but ultimately do not provide a long-term, reusable solution. In contrast, maturing technologies in data mining, computer vision, and machine learning, coupled with rapidly improving and affordable parallel and distributed hardware, have the potential to solve current and future image mining problems in an efficient, scalable way.

In this talk, we will provide an overview of our work toward systems and algorithms that can extract semantically meaningful content from data, with emphasis on specific algorithms that have been developed for visual recognition, querying, and discovery. We will also describe an exciting new project, the Autonomous Sciencecraft Constellation (ASC), that involves integrating perception, planning, and execution capabilities to create a highly capable spacecraft constellation that can make onboard decisions and carry out actions based on the content of the data collected.

Michael C. Burl received the PhD degree in Electrical Engineering from the California Institute of Technology in 1997 with a dissertation entitled "Recognition of Visual Object Classes". He is currently a Technical Group Leader and Senior Staff Member in the Machine Learning Systems Group at the Jet Propulsion Laboratory. He is the inventor of Diamond Eye, a distributed architecture for large-scale image mining and he also developed the core algorithms in JARtool, a tool for automatically cataloging volcanoes in the Magellan SAR imagery of Venus. He previously worked in the Battlefield Surveillance Group at MIT Lincoln Laboratory, where he developed algorithms for the detection and classification of tactical and strategic ground targets in high-resolution polarimetric SAR imagery. He is an organizer of the Third (and Fourth) Workshops on Mining Scientific Data Sets and has been a program committee member and invited speaker at various data mining and knowledge discovery conferences.

Hosted by Michael Mozer.
Refreshments will be served immediately following the talk in ECOT 831.

Department of Computer Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
May 5, 2012 (14:13)