Real-Time Face Mask Detection for Public Health and Safety Using a Lightweight MobileNetV2-Based Deep Learning Model
Keywords:
Face mask detection, Convolutional Neural Networks (CNN), MobileNetV2 architecture, Deep learning, Real-time image processing, OpenCV, Transfer learning, Data augmentation, Hyperparameter tuning, COVID-19 complianceAbstract
In our study, we tackle the pressing issue of monitoring face mask compliance during the COVID-19 pandemic by developing a real-time face mask detection system. Utilizing a combination of OpenCV for image processing and deep learning techniques, we create a reliable solution for differentiating between individuals wearing masks and those who are not. Our approach employs a Convolutional Neural Network (CNN) using the MobileNetV2 architecture, which we train on a custom dataset comprised of masked and unmasked images. With an impressive 98.2% accuracy on the training set and 97.3% on the test set, our model demonstrates its effectiveness. Additionally, the system is capable of processing video frames in real time and detecting multiple faces at once. We also explore various performance optimization strategies such as data augmentation, transfer learning, and hyperparameter tuning. Our face mask detection system has potential applications in access control systems, public transportation, and retail environments where ensuring mask compliance is essential.
