THE DEVELOPMENT OF FACE RECOGNITION MODEL IN INDONESIA PANDEMIC CONTEXT BASED ON DCNN AND ARCFACE LOSS FUNCTION

Abstract

The advancement of technology opens opportunities for implementation that benefits the social and economic aspects of human life. Given the latest achievement in face recognition technology that surpasses human ability to identify a face, the research explores the application of this scientific discovery in the Indonesian context during the current pandemic situation. Toward the effort to achieve this goal, the study develops an Indonesia Labelled Face in the Wild (ILFW) that collects face images of famous Indonesian people from the Internet in various poses, expressions, lighting/illumination, and fashion attribute. In response to the recent COVID-19 pandemic situation, the study also augmented a face mask to a portion of collected face images. Using DCNN, RetinaFace as the face detection model, and Arcface loss function, and adopting CRISP DM, the research contributes by providing a method to develop a face dataset with 1,200 identities, and face recognition model with 92 percent accuracy and be able to recognize Indonesian people with a face mask. The researchers also recommend use cases for realtime face recognition in the business organization. It uses CCTV to perform automatic attendance, security surveillance, and employee location tracking and exhibits deployment consideration. Future research could increase the accuracy of face recognition model by adding more identities to the face dataset.