Search

The global pandemic due to the novel corona virus, covid-19 has affected millions of lives across the globe. It has also disturbed economy, environment and social norms leading to many problems and giving birth to different rules and laws in order to ensure public safety. Wearing masks is one of the most important and primitive precautionary measures along with safe social distancing as advised by the World Health Organization. To manually monitor people not wearing face mask in order to ensure public safety is definitely a strenuous task. Therefore, this research work proposes a real time face mask detection system by applying computer vision and machine learning concepts like convolution neural networks and refined MobileNetV2 architecture to ease the deployment of proposed model in embedded devices with limited computational capacity. The dataset utilized here is available on Kaggle as Face mask detection dataset. The model is trained using Adam Optimizer algorithm which is best suited for deep learning models and is built using Keras, TensorFlow and OpenCV. The proposed model touches 99% accuracy under various training to testing ratios like 70% training and 30% testing,50% training and 50% testing etc. Precision, recall, f-score and support are calculated for all trials. This means that the system is computationally effective and could potentially be used in places like railway stations, airports or any other public places to detect people not wearing face mask and ensure safety to certain extent during this pandemic times.