Comparative Analysis of Object Detection Models for the Detection of Multiple Face Masks

ABSTRACT

Deep learning has immense prospective in many real-life practices, one of them being object detection. Object detection based on deep learning has shown encouraging results. Since December 2019, deadly virus named CORONA or COVID-19 started to engulf the whole planet with its impact. One of the easiest and simplest ways to protect oneself from this virus is by wearing a mask. In order to detect whether a person is wearing mask or not, we propose a model to detect various face masks that include cloth masks, N-95 masks, medical masks, and no mask. The proposed model consists of two major components—annotating, labeling images and detection of face masks. A new dataset has been created by combining images from Medical Masks Dataset and Google Images, and then these images were annotated according to the mentioned categories. A comparative study has been presented among different object detection algorithms along with a proposed detection algorithm. Results show that YOLOv5 performs best in the detection of face masks when compared to other detection models. It achieved a mAP of 0.51 in just 0.24 h on our dataset. On comparing YOLOv5 to the proposed model, we found that our model achieved a precision of 0.9 as compared to 0.88 of YOLOv5. Among existing approaches YOLOv5 performed the best with precision of 0.88. The model proposed in the work results in precision of 0.90 outperforming all existing models. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.