African Gender Classification Using Clothing Identification Via Deep Learning

Abstract

Human attributes identification and classification are popular aspects of computer vision which have been utilized in building relevant innovative systems in recent years. Most of these systems heavily rely on detection and recognition of facial attributes to perform efficiently. This work explores the use of an alternative approach to gender attribute classification of Africans by identifying traditional attires. Traditional attire is very popular among African societies and reflects attributes such as ethnic background, social status, and gender. The gender attributes of Africans reflected by the visual information of African traditional attires are explored in this work for gender classification. The AFRIFASHION1600 dataset was used to train a deep learning model, which achieved 87% accuracy on the test set.

Tasks

Reproductions