Single- and cross- database benchmarks for gender classification under unconstrained settings

Gender classification is one of the most important tasks in automated face analysis, and has attracted the interest of researchers for years. Up to now, most gender classification approaches have been tested using single-database experiments, and on quite controlled datasets such as the FERET database, which are not representative of real world settings. However, a recent trend towards more realistic benchmarks has emerged within the face analysis community, leading to the appearance of databases and protocols such as the Labeled Faces in the Wild (LFW) database, and the so-called Gallagher´s database, which comprises images collected from Flickr. Contrary to LFW, where a standard protocol for gender classification has been established as one of the BeFIT challenges, there is no standard protocol in Gallagher´s dataset, and a key contribution of this paper is to propose a standard 5-fold cross validation protocol for this database. Moreover, we provide cross-database experiments between Gallagher and LFW, as a way of assessing the performance of proposed algorithms in realistic conditions. In addition, we revisit and compare appearance-based (pixels) and feature-based (Gabor and LBPs) descriptors combined with linear SVM-based and LDA-based classification, carrying out single-database (LFW and Gallagher´s) and cross-database (Gallagher´s → LFW and LFW → Gallagher´s) experiments using the existing BeFIT challenge and the proposed dataset and protocols.