Speakers: Marcos A Rodrigues and Mariza Kormann - GMPR-Geometric Modelling and Pattern Recognition Research Group, Sheffield Hallam University

Marcos has published over 160 research articles in international journals, conference proceedings and book chapters, and have been awarded over 20 research grants and contracts from EPSRC, LSI, MRC, JISC, EU, industry and charity mainly on the subjects of robotics and AI, advanced modelling, 3D imaging, machine learning and pattern recognition. He is the Head of the GMPR-Geometric Modelling and Pattern Recognition Research Group, with research focused on pattern recognition and sensor design for a wide variety of applications ranging from robotics and automation, medical engineering, security, games and animation. He has developed and been awarded several patents on world-leading technology for fast 3D acquisition and reconstruction.

Mariza has been awarded an MSc with Distinction in Landscape Archaeology from the University of Oxford in 2009. Her dissertation work focused on 3D spatial modelling in connection with ancient landscapes, with particular reference to the case study of the Helike Plain in the Northern Peloponnese, Greece. Mariza joined Sheffield Hallam University in 2011 as Research Associate on the JISC funded project 3D Scanning of Museums Sheffield Metalwork Collection. Subsequently she worked as Research Associate on the EU-funded MARWIN project from November 2011 until Oct 2013. She joined the EU-funded ADMOS project in September 2013 as a Research Associate. Apart from her work on ADMOS, Mariza has strong interests and is pursuing archaeological research in connection with 3D spatial modelling and visibility.

Title: Advanced Techniques for Automatic Gender Classification

ADMOS is an FP7-funded project on real time analytics whose aims are to determine the effectiveness of an advert (e.g. words and images on a poster) placed in public spaces. The SHU/GMPR component of the project is to develop a real-time face detection and tracking, automatic gender classification and age estimation. The seminar will discuss the issue of gender classification and present our most recent results (over 90% accuracy for both male and female subjects) using a combination of pattern recognition techniques namely Local Binary Patterns, Eigenvector Decomposition, Census Transform, Discrete Cosine Transform and Support Vector Machines.

Please email Rachel Finch to book your place.