Title: Current Work in Computer Vision and Large Data Analysis at Kitware Inc.

Speaker: Will Schroeder, Anthony Hoogs, and Berk Geveci, Kitware, Inc.

Date/Time: Thursday, January 20, 2011, 11:00 am – 12:00 pm       

Location: CSRI Building/Room 90 (Sandia NM)

Brief Abstract: Kitware researches, develops and deploys open source toolkits and applications such as VTK and ParaView, which are in wide use in the scientific computing communities. In this presentation we will describe recent efforts in the areas of visualization, informatics and computer vision. In particular, we will describe our efforts visualizing on mobile devices, analyzing document repositories, and streaming large data. We will also describe our recent efforts in computer vision addressing wide-area video, an area for which Kitware has been recognized due to the recent award of two large DARPA grants.

Abstract: Video Analysis at Kitware: Events, Anomalies and Activities
Full-motion aerial video has radically changed the nature and practice of intelligence collection and processing in the US military over the past 10-15 years. Recently, another paradigm shift in video-based intelligence has been introduced with wide-area video, which can image an entire city at once. Both forms of video collect many more pixels than can be analyzed or even viewed by intelligence analysts, and consequently much effort has been expended on automating video processing. Historically the research emphasis has been on automated detection and tracking of vehicles, but the analyst is primarily interested in human behaviors, which are only partially evident in vehicle tracks. At Kitware we have been performing research in detecting and classifying human actions, events and activities in video. On the DARPA VIRAT program, we have led the development of a system that detects and tracks all people and vehicles on full-motion video such as Predator; characterizes their behavior in the form of actions of events; indexes this information; and retrieves archive videos showing behaviors similar to a query video. On the DARPA PerSEAS program, we are developing a system to detect emerging large-scale activities and threats, such as insurgent attacks, before they are carried out through online analysis of wide-area video. The PerSEAS system incorporates VIRAT content descriptors, as well as second-order anomaly detection, activity recognition and functional analysis. This talk will describe these programs and highlight a few of the underlying technologies in more detail.

Biography: Dr. Schroeder is President, CEO and co-founder of Kitware, Inc. Will's role at Kitware is to identify technology and business opportunities, and to obtain the necessary support for Kitware to meet these opportunities. Dr. Schroeder provides technical leadership in projects such as NLM's Insight Toolkit project (itk.org), NSF-funded projects for higher-order finite element visualization and visual databases, DOE grid computing, and various projects for medical image analysis.

Biography: Dr. Hoogs joined Kitware in August 2007 and founded a computer vision group that now has more than 15 members, most with PhD’s. Over the past 17 years, he has supervised and performed research in various areas of computer vision including: event, activity and behavior recognition; motion pattern learning and anomaly detection; tracking; visual semantics; learning, image segmentation; object recognition; and content-based retrieval. He has led research projects, sponsored by DARPA, AFRL, CIA, NGA and corporate funds, that developed advanced prototypes and demonstrations installed at operational facilities.

Biography: Dr. Geveci leads the scientific visualization and infomatics teams at Kitware Inc. He is one of the leading developers of the ParaView visualization application and the Visualization Toolkit (VTK). His research interests include large scale parallel computing, computational dynamics, finite elements and visualization algorithms.
Dr. Geveci regularly publishes and teaches courses at conferences including IEEE Visualization and Supercomputing conferences.

CSRI POC: David Rogers, 505-844-5323