Face Vision

Vision uses a structured output classifier based on the models of deformable parts. The quality of the configuration of the reference points for the given image is measured. The maximization of f is resolved by the dynamic programming, thanks to the form of the constraints of the graph (direct acyclic graph). The accuracy of the detector is measured in terms of relative deviation defined as the distance between the positions of the reference point of the estimated truth and divided by the size of the face. The size of the face is defined as the distance between the center of the mouth and the midpoint between the eye centers. Geometrical accuracy is measured by the average displacement of the relative characteristic and by the relative maximum displacement of the characteristic to which it is measured in terms of deviation, relative, defined as the distance between the positions of the estimated reference point and divided by the face size. The size of the face is defined as the distance between the center of the mouth, and the midpoint between the eye centers. Geometrical accuracy is measured by the average displacement of the relative characteristic and the relative maximum displacement of the characteristic and is compared with three concurrent detectors in terms of accuracy of the estimated positions. Specifically, it is compared with the parameters based on the following approaches: active appearance models, detectors based on deformable models and binary formed independently for each reference point.