Detecção de pontos fiduciais em faces humanas utilizando máquina de vetores suporte

Abstract

Currently, the problem of detention fiducial points on human faces has received increasing attention from the scientific community. Recently, we may find in the literature developed some systems for the purpose of performing detecçiia fiducial points. Some of these representatives use clasSifieadows SVM. One of the problems faced by this. fano approach is that the performance of SVM classifiers is extremely sensitive to change its parameters. list is a difficulty that is not trivial to exploit. In this lecture tion, we propose a fiducial points detection system, which uses SVM classifiers, in order to investigate the performance of the classifiers for con-along pre-defined parameters. With this investiwida, we intend to describe the system's behavior to the set of parameters and determine what the parameters = bination, and its values, which produces the best pos-sible performance for the branch system a whole. The proposed system is complex due to. lot of steps involved. The measured performance of the proposed system was performed. for. eleven Aduri tips: ais of. human face. two bases of image data (BiolD and End). The results demonstrate that the system performance. proposed is incremented when using the approach adopted. The system performance is satisfactory when compared with similar systems.