In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
will defend her dissertation proposal
Towards Improving Set-Based Face Recognition System
The face recognition system allows us to distinguish individuals easier, faster, and in a more secure manner. In a set-based face recognition system, multiple images from the same individuals form an image set and are employed to enhance the recognition performance. However, despite the encouraging results obtained, the set-based face recognition remains a challenging problem in the real-life scenario. Regarding the comparison objects, we consider two situations: (i) set-to-set matching, and (ii) point-to-set matching. In set-to-set matching, the comparison happens between two image sets. The major challenge is to generate efficient and robust signatures with identity information from an image set with noise, outliers, variations, etc. In point-to-set matching, the comparison happens between an image set and a single image. Since the image sets and images are usually represented with different models, their signatures share different statistical properties. The major challenge is to adapt them to the same space so that they are comparable. The goal of this research is to develop new signature generation and matching algorithms that increase the face recognition performance of set-based systems. The specific objectives are to develop and evaluate: (i) a signature generation and similarity measurements algorithm for set-to-set matching, (ii) a signature compression and binarization algorithm for set-to-set matching, and (iii) a signature generation algorithm for point-to-set matching. The proposed algorithms achieved significant improvements when compared with previous baselines.
Date: Monday, March 19, 2018
Time: 10:45 AM
Place: HBS 317
Advisors: Dr. Ioannis A. Kakadiaris
Faculty, students, and the general public are invited.