A Novel Approach for Detecting Face Masks and Social Distancing in Public Places

A Novel Approach for Detecting Face Masks and Social Distancing in Public Places

UshaRani Shola, Vishnu Vadan Mandapati, Gayathri R., Eswari Sudha Tummalacharla
Copyright: © 2023 |Pages: 22
DOI: 10.4018/978-1-6684-5741-2.ch016
OnDemand:
(Individual Chapters)
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

The system is fed with the CCTV footage or a real-time web camera, where the people in the frame get detected and the face mask detection module takes place. It uses the facial recognition and identifies a person without mask, followed by a notification alert message through a mail. Then the social distancing module checks whether the person is in a safe zone by measuring social distancing with nearby people. It alerts as “Please put on the mask. The people near you are not following social distancing.” The CCTV cameras are set-up in places where the system needs to follow the above violation metrics. Extra CCTV cameras can be installed if certain areas are meant to be monitored where CCTV cameras were not previously installed. Then the servers or computers must be installed to run inference of the received footage. The system uses a deep learning model in the computer servers to identify the above specified violation metrics.
Chapter Preview
Top

Background Study

This work in (Jiang et al., 2020) proposes a two-stage CNN (Chavda et al., 2021) architecture, with the first stage recognizing human faces and the second stage employing a lightweight image classifier to classify the faces detected in the first stage as 'Mask' or 'No Mask' faces and draw bounding boxes around them with the observed class name. The face detector extracts all of the faces in the image and outputs them (Ao et al., 2009) together with their bounding box coordinates.

ResNet is the normal backbone in RetinaFaceMask (Punn et al., 2020), but MobileNet is offered for comparison and to reduce computation and model size in deployment circumstances with restricted computer resources. FPN is used as a neck in RetinaFaceMask, and It may extract high-level semantic information and then utilise an addition operation with a coefficient to merge it into the feature maps of previous layers. Because it can have varied receptive fields to detect varying sizes of objects, RetinaFaceMask uses a multi-scale detection technique similar to SSD to create a prediction with several FPN feature maps.

Complete Chapter List

Search this Book:
Reset