[Paper Review] Social Distance Detection Using Deep Learning And Risk Management System

R. G. Sangeetha, Jaya Aravindh V. V|arXiv (Cornell University)|Apr 20, 2023

COVID-19 diagnosis using AI4 citations

TL;DR

This paper proposes a deep learning-based social distance detection system using single-stage object detection with CNNs to monitor compliance in real time via CCTV footage. It integrates a risk management module to analyze violations, enabling automated surveillance in public spaces like hospitals and schools, reducing reliance on manual monitoring while improving safety during pandemics.

ABSTRACT

An outbreak of the coronavirus disease which occurred three years later and it has hit the world again with many evolutions. The effects on the human race have already been profound. We can only safeguard ourselves against this pandemic by mandating a "Face Mask" also maintaining the "Social Distancing." The necessity of protective face masks in all gatherings is required by many civil institutions in India. As a result of the substantial human resource utilization, personally examining the whole country with a huge population like India, to determine whether the execution of mask wearing and social distance maintained is unfeasible. The COVID-19 Social Distancing Detector System is a single-stage detector that employs deep learning to integrate high-end semantic data to a CNN module in order to maintain social distances and simultaneously monitor violations within a specified region. By deploying current Security footages, CCTV cameras, and computer vision (CV), it will also be able to identify those who are experiencing the calamity of social separation. Providing tools for safety and security, this technology disposes the need for a labor-force based surveillance system, yet a manual governing body is still required to monitor, track, and inform on the violations that are committed. Any sort of infrastructure, including universities, hospitals, offices of the government, schools, and building sites, can employ the technology. Therefore, the risk management system created to report and analyze video streams along with the social distance detector system might help to ensure our protection and security as well as the security of our loved ones. Furthermore, we will discuss about deployment and improvement of the project overall.

Motivation & Objective

• To address the challenge of manual monitoring of social distancing in high-population areas like India.
• To reduce human resource dependency in enforcing public health measures during the COVID-19 pandemic.
• To develop an automated system that detects violations of social distancing and mask-wearing using computer vision.
• To integrate a risk management system for reporting and analyzing video-based violations.
• To enable scalable deployment across institutions such as schools, hospitals, and government offices.

Proposed method

• The system uses a single-stage detector with a CNN-based architecture to extract high-level semantic features from video frames.
• It processes real-time video streams from existing CCTV cameras to detect individuals and measure inter-personal distances.
• The system applies deep learning models trained on person detection and spatial relationship understanding to identify proximity violations.
• A risk management module logs and analyzes detected violations, enabling tracking and reporting for administrative follow-up.
• The framework is designed for integration into existing surveillance infrastructure without requiring new hardware.
• It supports deployment in diverse environments such as universities, hospitals, and construction sites.

Experimental results

Research questions

• RQ1 How can deep learning be used to automate social distance monitoring in public spaces?
• RQ2 What is the effectiveness of a CNN-based single-stage detector in detecting proximity violations in real-time video?
• RQ3 How can a risk management system enhance the utility of automated detection for institutional enforcement?
• RQ4 Can the system be deployed effectively in diverse public environments with minimal infrastructure changes?
• RQ5 What is the potential reduction in manual surveillance workload through automation?

Key findings

• The system enables real-time detection of social distancing violations using existing CCTV infrastructure, reducing manual monitoring needs.
• The integration of a risk management system allows for structured logging and analysis of detected violations.
• The model is deployable across various public institutions, including schools, hospitals, and government offices.
• The approach supports simultaneous detection of mask-wearing and social distancing, enhancing public health compliance.
• The system demonstrates scalability and adaptability to different environments with minimal hardware upgrades.
• The solution reduces operational costs and increases consistency in enforcing public health measures.