Mixture of experts for classification of gender, ethnic origin, and pose of human faces

Abstract

In this paper we describe the application of mixtures of experts on gender and ethnic classification of human faces, and pose classification, and show their feasibility on the FERET database of facial images. The FERET database allows us to demonstrate performance on hundreds or thousands of images. The mixture of experts is implemented using the "divide and conquer" modularity principle with respect to the granularity and/or the locality of information. The mixture of experts consists of ensembles of radial basis functions (RBFs). Inductive decision trees (DTs) and support vector machines (SVMs) implement the "gating network" components for deciding which of the experts should be used to determine the classification output and to restrict the support of the input space. Both the ensemble of RBF's (ERBF) and SVM use the RBF kernel ("expert") for gating the inputs. Our experimental results yield an average accuracy rate of 96% on gender classification and 92% on ethnic classification using the ERBF/DT approach from frontal face images, while the SVM yield 100% on pose classification.

Mood-Learning Public Display: Adapting Content Design Evolutionarily through Viewers' Involuntary Gestures and Movements

“…There are many studies on obtaining information through the viewer's appearance, especially through face detection [5], [6], and many products for public display are commercially-available. These R&Ds enable the public display system to get viewers' demographic segmentations, such as sex and age, while the information does not include viewers' attitudes towards public display, and thus the system cannot estimate what viewers really want to see.…”

Face Gender Classification on Consumer Images in a Multiethnic Environment

“…But different from gender classification, ethnicity classification is much harder and sometimes even human can not have a very clear division for ethnicity in perception. In literature, G. Shakhnarovich et al [8] divided ethnicity into two categories: Asian and Non-Asian, while in [7] [18] [19] three categories with Mongoloid, Caucasoid and African were adopted, and in [17] four ethnic labels with Caucasian, South Asian, East Asian, and African are used. In this paper, we use three ethnic labels with Mongoloid, Caucasian and African.…”

Gender Classification with Bayesian Kernel Methods

“…[3] used a neural network and showed that even very low resolution image such as 8x8 can be used for gender classification. [4] used the mixture of experts with ensembles of RBF networks and a decision tree as a gating network. [5] showed that SVMs worked better than other classifiers such as ensemble of RBF networks, classical RBF networks, Fisher linear discriminant, nearest neighbor etc.…”