ADVANCED FACE MASK DETECTION USING TRANSFER LEARNING AND CUSTOM CLASSIFIERS: ENHANCING PUBLIC SAFETY THROUGH COMPUTER VISION AND DEEP LEARNING

Face masks offer protection against air pollution and spread of the disease and should be worn for effortful distancing. Video cameras have proven useful upholding uniformity of mask-wearing using computer vision. Earlier mentioned methods based on convolutional neural network (CNN), YOLO (you only look once) and faster R-CNN support vector machines (SVM), and haar cascade techniques have had difficulties mainly for frontal view faces. This study provides nascent developments that will bring about remarkable changes in the field of public health as well as technologies with a unique mask detection system based on the latest computer vision combined with deep learning medicine. Using transfer learning and mobileNet V2, custom 'faceNet' and 'MaskNet' classifiers were run, which resulted in the 98,3% accuracy of the model - 97.87% for women with masks and 98,46 % for women without masks. This technology, in addition to mask effectiveness monitoring, can upgrade the approach in the CCTV monitoring process from performing passive surveillance to facilitating and enabling new technology that enhances safety and health practices.