Preserving Women Public Restroom Privacy using Convolutional Neural Networks-Based Automatic Gender Detection

Abstract

Personal safety and privacy have been the significant concerns among women to use and access public restrooms/toilets, especially in developing countries such as Indonesia. Privacy-enhancing designs are unquestionably expected to ensure no men entering the rooms neither intentionally nor accidentally without prior notice. In this paper, we propose a facial recognition approach to ensure women's safety and privacy in public restroom areas using Convolutional Neural Networks (CNN) model as a gender classifier. Our main contributions are as follows: (1) a webcam feed automatic gender detection model using CNN which may further be connected to a security alarm (2) a publicly available gender-annotated image dataset that embraces Indonesian facial recognition samples. Supplementary Indonesian facial examples are taken from a government-affiliated college, Politeknik Statistika STIS students' photo datasets. The experimental results show a promising accuracy of our proposed model up to 95.84%. This study could be beneficial and useful for wider implementation in supporting the safety system of public universities, offices, and government buildings.