Der Rückblick zum TEWI-Kolloquium von Judith Redi, TU Delft am 23.11.2012 beinhaltet die Folien:
Abstract: Machine learning has been recently shown to be a very promising tool to support automated QoE assessment. Its ability to mimic highly non-linear, complex phenomena, such as user experience and quality judgment, is extremely appealing for the implementation of on-line, accurate QoE control systems. Nevertheless, applying Machine Learning to QoE is an high risk, high gain approach: if misused, it can lead to poorly flexible and unreliable systems. Key to the attainment of all the gain without risks is a profound understanding of the advantages and limitations that characterize learning machines. Even more important is a strong knowledge of the phenomenon to be mimicked, that is, the user experience.
In this talk, theoretical background, applicative tips and practical examples will be reviewed, with the aim of drawing guidelines for the successful application of Machine Learning to objective QoE assessment.
CV: Judith Redi is Assistant Professor at Delft University of Technology, department of Intelligent Systems, since 2010. She obtained her PhD from the University of Genoa (Italy) in 2010, with a thesis on learning machines for objective image quality assessment, final result of a project on
visual quality in displays funded by Philips research. After receiving the award for the best ICT thesis from University of Genoa, she worked as a Post-Doc at Eurecom (France) focusing on image analysis and computer vision. At TU Delft, she works on image and video understanding towards the maximization of the quality of multimedia experiences, for which she was awarded an NWO Veni grant in 2012. She is coordinator of the Qualinet (COST IC1003) Industrial Forum and Management Committee member for the Qualinet COST action.