Through connectivity in applied computer systems – ADMOS and MARWIN projects

RODRIGUES, Marcos and KORMANN, Mariza (2015). Through connectivity in applied computer systems – ADMOS and MARWIN projects. In: 5th Annual International Scientific Conference on Education, Science, Innovations ESI 2015, Pernik, Bulgaria, June 10-11, 2015. (Unpublished)

[img]
Preview
PDF (FP7 ADMOS and MARWIN Projects)
Rodrigues_ESI_presentation.pdf - Presentation
Available under License All rights reserved.

Download (31MB) | Preview

Abstract

The ADMOS project is funded by the EC and aims to develop a real-time gender classification and age estimation system to be used in private spaces for public use, such as shopping malls, fairs and outdoor events. The purpose is thus, the collection of demographic statistics for marketing purposes including the development of a customised human-computer interface. On top of requirements for real time processing, tracking faces in a public space raises serious privacy issues. A video frame contains images of faces, which by definition, are classed as personal data. The European Union has clear regulations on the acquisition and processing of personal data, which we are obliged to comply with. In this paper, we discuss how issues of fast computation in real time were solved through a combination of LBP-Local Binary Patterns and SVM-Support Vector Machines. Furthermore, we discuss how we were able to track anonymous faces over multiple frames using a method that satisfy EU regulations on privacy. Our solution uses a system of multi-level queues similar to multi-level feedback queues in operating systems. In recent review of the project in Brussels the reviewers suggested that our solution has the potential to set standards for future projects on face detection in public areas.

Item Type: Conference or Workshop Item (Keynote)
Research Institute, Centre or Group: Cultural Communication and Computing Research Institute > Communication and Computing Research Centre
Depositing User: Marcos Rodrigues
Date Deposited: 18 Jun 2015 12:37
Last Modified: 20 Aug 2015 11:49
URI: http://shura.shu.ac.uk/id/eprint/10292

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year

View more statistics