On stream face mask and helmet detection system using YoloV3 algorithm

Safety and hygiene are the primary criteria to ensure a healthy and safe human survival in the present-day scenario. With the devastating Coronavirus pandemic research has consistently shown that basic hygiene is necessary, wearing a mask in public places is essential. Also considering wearing a helmet as a safety measure to stringent for two-wheeler riders. A real-time yolov3 object detector is used to train the model to detect facial masks and helmets in video footages, live feeds or images to look for public hygiene and safety. It employs a neural network for prediction bounding boxes and their possibilities throughout the entire image. The dataset has been obtained online. The model has been trained using online images for different classes like wearing mask, no helmet; not wearing helmet and mask; wearing helmet, no mask; wearing both helmet, mask. The trained dataset is mainly used to classify and find out objects for the mentioned classes. If no helmet or mask is found, the SMTP library module provided by python describes a SMTP to send a mail to any internet device using SMTP or ESMTP as a client session object. The proposed system was put to the test on a dataset consisting of images of people following COVID-19 safety protocols and pictures of people wearing helmets following road safety guidelines. The results obtained have high accuracy and precision.