The COVID19 pandemic and the variants of COVID19 have affected millions of people globally. With no effective cure, prevention is the only solution to slow down the rapid transmission of the virus. Face masks play a vital role in prevention of the transmission. Real time detection of proper face masks is crucial in this time. The proposed system helps in identifying proper face mask automatically compared to the manual detection. We propose a Localized YOLO algorithm in this paper which identifies the proper face masks using following two criteria, 1) Nose, Mouth and Chin is fully covered 2) Wearing any other object to cover face is detected. The DarkNET is used to focus more on the important features of the image. The spatial difference between the prediction and ground truth boxes are lowered in this proposed method for increased precision in prediction by adopting the GloU loss method. We used the Kaggle Face Mask detection dataset containing 853 images, the data set is processed for three different categories namely 1)Without Mask, 2)With Proper Mask, 3)With Incorrect Mask. The proposed Localized YOLO has good prediction performance. The proposed algorithm can be used in real time over a video to detect people with incorrect face masks.
Key words: Shanmughapriya M, Brindha Devi V, Josephine Ruth Fenitha, Sanchana R
|
