Motion DSP - Video Examples. Background:Shoplifting and employee theft cost US businesses billions of dollars annually. Retailers maintain a database of known thieves and shoplifters, and keep photos of the individuals. They compare surveillance images with their photo database of known shoplifters, and if a match is found, they must quickly pass the image to local law enforcement so the suspected shoplifter can be apprehended before leaving the area. Speed is of the essence. The challenge is that surveillance images need to be high quality in order to be matched accurately to the database, and the quality of images from retail security cameras (also known as CCTV cameras) can vary significantly due to the camera’s location, resolution, and the movement of the individual.
MotionDSP’s Role:MotionDSP’s Ikena Forensic software is able to significantly increase the fidelity of images captured from video so the clearest image can be matched to the database and given to law enforcement. Data visualization algorithms. Assistant Professor Computer Science Purdue University My research is on high performance and large scale matrix computations and I tend to focus on analyzing large data from information networks, social networks, protein networks, and scientific simulations.I received an NSF CAREER award to discover how to scale matrix methods to the enormous sizes of modern data with only modest computational resources.
Prior to joining Purdue, I was the John von Neumann post-doctoral fellow at Sandia National Labs. Interests I wrote an article for ACM’s XRDS magazine on some of the issues that arise in my research. It’s a nice overview of issues I’m looking at and problems I’m studying. David F. Matrix-based network computations PageRank Spectral graph theory Parameterized models of physical systems Simulation data MapReduce and high performance computers Dynamic networks Community detection and clustering Large scale data computations Email dgleich--at--purdue.edu Classes Prior classes.
Sodaplay. Army of clerks. Craig W. Reynolds. My interests center around using procedural models (computer programs) to simulate complex natural phenomenon. These models can aide scientific understanding of the natural system. They also allow us to recreate the phenomenon and control it for use in animation, games and the arts. Much of my work involves writing software to simulate various types of human and animal behavior. These programs control the actions of autonomous characters in virtual worlds. I started by simulating bird flocks and related group behaviors. That approach was generalized to other kinds of goal directed steering behaviors. I am also interested in using evolutionary computation to design procedural models, such as for behavioral control and texture synthesis. I work as a software engineer at RightHook in San Jose, California. Transphormetic.