Place: Large Lecture Room
Affiliation: Electrical and Computer Engineering Department of the American University of Beirut
Using local invariant features has been proven by published literature to be a powerful tool for image processing and pattern recognition tasks. However, in energy constraint environments, these invariant features would not scale easily because of their computational requirements. Motivated to find efficient building recognition algorithms based on the scale invariant feature transform (SIFT) keypoints, we discuss in this talk results of uSee, a supervised learning framework that identifies subsets of relevant keypoints. With only 14.3% of an image SIFT keypoints, uSee exceeded prior literature and identified with an accuracy of 99.1% buildings in the Zurich Building Database (ZuBud).
Aside from the energy aware building recognition project, Dr. Awad will be also presenting some of her research work on biologically inspired deep belief networks, wild sport video analysis, augmented reality and interactive environments using mobile apps….