Smart Surveillance: Deep Learning-Based Real-Time Fire and Accident Detection

Abstract

In this paper fire detection system and real time accident using the YOLOv5 deep learning framework is proposed. In smart cities there is increasing need for intelligent surveillance which has driven the adoption of computer vision techniques to monitor and respond to critical events. In this work, YOLOv5 is trained on a dataset which includes annotated vehicular accident and fire scenarios to enable detection accurate across diverse conditions. The model is appropriate for live video monitoring applications because of its architecture which strikes a balance between accuracy and speed. The system reduces response time and facilitates decision-making for emergency management by automatically sending out notifications as soon as occurrence is recognized. Experimental findings gives system’s resilience for practical implementation which can achieve high detection accuracy with few false positives. The suggested framework demonstrates combining cutting-edge object identification models with surveillance system which enhance public safety, reduces losses and support the development of smart and sustainable urban infrastructures.