A Novel Genetic Algorithm for 3D Facial Landmark Localization

This paper introduces Evolutionary Putsuit (EP) as a novel approach for 3D facial landmark localization. Leveraging the global optimization ability of genetic algorithm (GA), an innovative framework for detecting local features, like nose tips and eye-corners, on 3D face data is proposed. Facial landmarks are localized by combining responses of several weak classifiers as basic building blocks. A genetic algorithm is employed for searching the optimal way to combine these building blocks, in order to construct a classifier with high detection accuracy. This GA-based facial landmark detection method is evaluated on the FRGC v1 database; it shows promising results that the GA searching process is able to produce an effective feature classifier from a relatively small training set, achieving a strong landmark localization ability.