In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

Nikolaos Sarafianos

will defend his dissertation proposal

Person Search Using Identity Attributes

Abstract

In cases of missing children or in surveillance applications there is a critical unmet need to go quickly through large volumes of data to identify the person of interest. Identity and identity-related attributes, such as soft-biometrics (gender, height), the type of clothing or the existence of accessories (backpacks, hats) can be a powerful representational approach to accomplish this task. Thus, if we could efficiently identify which images or videos contain humans with such characteristics we could reduce dramatically the labor and the time required to identify such traits. To accomplish this task, our goal is to develop and evaluate algorithms and a prototype system to search a video for frames depicting humans with specific identity attributes. Solving this problem requires addressing three separate subproblems, namely (i) defining the ontology of the identity and the identity-related attributes, (ii) developing and evaluating algorithms for extracting identity attributes from images, and (iii) developing and evaluating an algorithm for attribute-based person search in videos. In this proposal, we propose a series of methods to improve the accuracy of identity attribute classification systems. In each method a detailed overview of the benefits and limitations of each approach is introduced, extensive experimental evaluation and ablation studies are provided to analyze the impact of different modules, and further bottlenecks and limitations were identified that need to be addressed by remaining work.

Date: Friday, April 27, 2018
Time: 9:30 AM
Place: HBS 317
Advisors: Dr. Ioannis A. Kakadiaris

Faculty, students, and the general public are invited.