April 2017
Beginner to intermediate
250 pages
6h 28m
English
Modern face-tracking techniques are almost entirely data driven, that is, the algorithms used to detect the locations of facial features in the image rely on models of the appearance of the facial features and the geometrical dependencies between their relative locations from a set of examples. The larger the set of examples, the more robust the algorithms behave, as they become more aware of the gamut of variability that faces can exhibit. Thus, the first step in building a face-tracking algorithm is to create an image/video annotation tool, where the user can specify the locations of the desired facial features in each example image.
Read now
Unlock full access