Face Age Classification on Consumer Images with Gabor Feature and Fuzzy LDA Method

Abstract

As we all know, face age estimation task is not only challenging for computer, but even hard for human in some cases, however, coarse age classification such as classifying human face as baby, child, adult or elder people is much easier for human. In this paper, we try to dig out the potential age classification power of computer on faces from consumer images which are taken under various conditions. Gabor feature is extracted and used in LDA classifiers. In order to solve the intrinsic age ambiguity problem, a fuzzy version LDA is introduced through defining age membership functions. Systematic comparative experiment results show that the proposed method with Gabor feature and fuzzy LDA can achieve better age classification precision in consumer images.

Contourlet appearance model for facial age estimation

“…Sample images from the two databases are shown in Figure 6. The Mean Absolute Error (MAE), defined as the average of the absolute error between the es- timated ages and the ground truths, is computed using (11) and is used as our performance metric to compare the dif ferent age estimation techniques.…”

“…Moreover, skin zones such as wrinkles and blobs are also considered in this graph. Exploiting the strength of Gabor features, Gao and Ai [11] used them for facial image representation and used a Linear Discrim inant Analysis (LDA) for building an age classifier. They also proposed using fuzzy LDA to cope with age ambigu ity.…”

D2C: Deep cumulatively and comparatively learning for human age estimation

“…Typical features designed specifically for 70 age estimation include facial features and wrinkles [7], the learned AGES (AGing pattErn Subspace) [8] features, as well as the biologically inspired features (BIF) [13]. Other more general features devised for texture description are also widely used for age estimation, for example the LBP feature [5,14], the Gabor feature [6], etc.…”

“…Widely used features include local binary pattern (LBP) [5] and Gabor features [6], with some further processing models like the anthropometric model [7], AGing pattErn Sub-10 space (AGES) [8], and the age manifold model [9]. To learn an age estimator, most approaches use either a multi-class classification framework or a regression framework.…”

“…Classification models, e.g. linear SVM [13], Probabilistic Boosting Tree [15], Fuzzy LDA [6], or regression models like Support Vector Regression [13], Kernel Partial Least Squares [16], Neural Network [17] and 80 Semidefinite Programming [18] are all designed to estimate age.…”