Performance Evaluation of CNN Models for Face Detection and Recognition with Mask

Abstract

Face recognition is widely used across many biometric applications. In recent times, where COVID-19 has impacted manufacturing, travel, tourism, hospitality, and crippling the global economy, wearing a mask is necessary for many establishments and public places for its widespread and safety of an individual. Under this scenario, person recognition for security functions is been difficult for the present face recognition systems in which most of the facial features are covered. Our work relies on computer vision and deep learning models which intend to make an impact and solve the real-world problem of safety measures at some significant level. In our work, we propose a framework for masked face recognition using Inception v3 and FaceNet architectures, which can be easily integrated into various embedded devices with limited computational capacity. We aim to detect the face with a mask and recognize the person in images as well as in real-time videos. We demonstrate the results with an overall accuracy of 88% for masked face recognition within the defined scope.