ABSTRACT
The COVID-19 pandemic has infected tens of millions of individuals around the world, and it is currently posing a worldwide health calamity. Wearing a face mask in public places is one of the most effective protection strategies, according to the World Health Organization (WHO). Moreover, their effectiveness has declined due to incorrect use of the face mask. In this scenario, effective recognition systems are anticipated to ensure that people's faces are covered with masks in public locations. Many people do not correctly wear the masks due to inadequate practices, undesirable behaviour, or individual vulnerabilities. As a result, there has been an increase in demand for automatic real-time face mask detection and mask position detection to substitute manual reminders. This proposed work classifies people into three categories such as with mask, without mask and mask with incorrect position. The dataset is tested using three different variants of object detection models, namely YOLOv4, Tiny YOLOv4, and YOLOv5. The experimental result shows that YOLOv5 model outperforms with the highest mAP value of 99.40% compared to the other two models. Copyright © 2023 Inderscience Enterprises Ltd.
ABSTRACT
The COVID-19 pandemic has infected tens of millions of individuals around the world, and it is currently posing a worldwide health calamity. Wearing a face mask in public places is one of the most effective protection strategies, according to the World Health Organization (WHO). Moreover, their effectiveness has declined due to incorrect use of the face mask. In this scenario, effective recognition systems are anticipated to ensure that people's faces are covered with masks in public locations. Many people do not correctly wear the masks due to inadequate practices, undesirable behaviour, or individual vulnerabilities. As a result, there has been an increase in demand for automatic real-time face mask detection and mask position detection to substitute manual reminders. This proposed work classifies people into three categories such as with mask, without mask and mask with incorrect position. The dataset is tested using three different variants of object detection models, namely YOLOv4, Tiny YOLOv4, and YOLOv5. The experimental result shows that YOLOv5 model outperforms with the highest mAP value of 99.40% compared to the other two models.
ABSTRACT
The global economy has been affected enormously due to the spread of coronavirus (COVID-19). Even though, there is the availability of vaccines, social distancing in public places is one of the viable solutions to reduce the spreading of COVID-19 suggested by the World Health Organization (WHO) for fighting against the pandemic. This paper presents a YOLO v3 object detection model to automate the monitoring of social distancing among persons through a CCTV surveillance camera. Furthermore, this research work used to detect and track the person, measure the inter-person distance in the crowd under a challenging environment which includes partial visibility, lighting variations, and person occlusion. Moreover, the YOLO V3 model experiments with Darknet53 and ShuffleNetV2 backbone architecture. Compared with Darknet53 architecture, ShuffleNetV2 achieves better detection accuracy tested on Custom Video Footage Dataset (CVFD), Oxford Town Centre Dataset (OTCD), and Custom Personal Image Dataset (CPID) datasets.