Face mask detection

• Main Author: Chua, Delise Yun Jie
• Other Authors: Fan Xiuyi
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/166133
• Institution: Nanyang Technological University

The sudden outbreak of the Covid-19 pandemic globally has disrupted the daily lives of millions of people, forcing them to adapt to the “new normal”, which includes mandatory wearing of facemasks and face shields in public areas. Amidst the new regulations, adapting to these new changes and attempting to form new habits still requires substantial time. Deployment of a face mask detector is thus essential to ensure compliance to the enforced guideline. In this paper, we explore three different models with Convolutional Neural Network (CNN) based architecture for detecting facemasks. Supported by the Haar cascade classifier, the best performing model will then be integrated with the Raspberry Pi 4 and built in webcam to perform real time facemask detection. Results from this study shows a promising outlook in terms of accuracy in detecting facemask with the 3 different Convolutional Neural Network architectures. Across the three models, it was observed that a trade-off occurs between accuracy and computation time. Performance of the system on the built-in webcam appears to be better than the Raspberry Pi in terms of accuracy and computation time. However, further refinements and finetuning remains necessary to enhance the performance during the real time implementation.