Detecção de Pontos Fiduciais sobre a Face em Tempo Real

Recently, considerable attention has been given to the problem of robust detection and tracking of facial features. In this work, a framework for facial landmarks detection is presented. This framework is comprised by three interchangeable parts, namely preprocessing, classification and post-processing. In the classification stage, we used a cascade of classifiers based on correlation filters called IPD (Inner Product Detector). The resulting system is able to recognize facial landmarks on different subjects in real-time, under various lighting conditions. Details such as training of the detector employed, validation and usage are also discussed.