Work of Lanitis et al
Work of Lanitis et al. used the Active Appearance Model (AAM), a statistical face model, to study age estimation problems. In their approach, after AAM parameters were extracted from face images landmarked with 68 points, an “aging function” was built using Genetic Algorithms to optimize the aging function. Meanwhile, Geng et al. introduced an AGing pattern Subspace (AGES) to estimate the ages of individuals. The AGES method models a sequence of individual aging face images by learning a subspace representation in order to handle incomplete data such as missing ages in the training sequence. For a previously unseen face image, its proper aging pattern is defined by the projection in the subspace with a minimum reconstruction error that recreate the face image, while its age is determined by the position of the face image in that aging pattern.