On Thursday, 2017-05-11, at 14:15 at the Faculty of Computer and Information Science, Vitomir Štruc will talk about pose variability representing one of the major problems not satisfactory addressed yet by the existing face-recognition technology. Recent advancements in facial landmark localization and image alignment have made it possible to devise new methods and techniques to mitigate this problem and successfully match facial images captured under significantly different viewpoints (or poses). In this talk, we will discuss some of the methods that can be used to address pose variability in facial recognition systems and present our recent work on (a) facial landmark detection from 2D and 3D data, (b) face frontalization, and (c) reposing of facial images. We will elaborate on some of the recent trends in the area of facial alignment and highlight other application areas where the presented techniques can be used.
SLAIS © 2015 Prijava