Deep transfer learning based real time face mask detection system with computer vision

The COVID -19 pandemic has great impact in almost all life in the world. Till a immunogen is identified, everyone must be very careful about the spread of corona virus. Wearing a mask will definitely prevent us from corona virus. This results from the requirement to keep everyone safe from the virus's propagation. However, because current systems can't more effectively match faces with masks, the steps made to stop the virus' propagation present a problem for security and surveillance systems. Three methods are proposed to identify or detect whether person is sporting a mask or improper wearing of mask or a person with no mask. The ResNet152V2,VGG16 will work with still pictures and jointly works with a live video stream. A mask detection dataset consists of pictures of both mask wearing and without mask wearing. OpenCV is used to do the real time face detection. It checks whether mask is properly worn over the nose area by completely covering chin and mouth. The proposed mask symbol is least complicated in structure and provides fast results so it is used in video surveillance. Mass screening is possible and thus utilized in thronged places like railway stations, bus stops, markets, streets, mall entrances, schools, colleges and so on. In proposed system, ResNet152V2 model is used for robust face detection. The head based model is selected for classifying faces as masked or non-masked. Finally computer vision concepts are added to improve performance on video streams. The results demonstrate that the proposed model improves accuracy by 99.1%, performance of precision by 99.2%, F1-score by 99.1%. It is identified that the proposed model ResNet152V2 accomplished highest results.