Abstract
Gender identification is a precursor for context specific emotion recognition. Small but significant differences have been noted across different gender groups in terms of emotion expressiveness. Apart from facial expressions and security applications, gender recognition is becoming increasingly relevant after the rise of applications involving social media platforms. Labelled Faces in the Wild (LFW) dataset is designed for studying the problem of face recognition under unconstrained environment. However, it is used to study other facial attributes as well, including gender. In this paper, we propose a standardized subset of LFW database (LFW-gender) that can be used as a benchmark for gender recognition algorithms. We also provide a baseline for performance on the dataset for gender recognition with various algorithms and some results may suggest that this is a harder subset to classify.