Automated Hazard Detection and Safety Monitoring for Smart Construction Using Deep Learning Algorithm

A construction site monitoring system combines several tools and technology to allow for remote monitoring of the site. The primary goal of this system is to maintain worker safety and to keep the project on schedule. This study proposes a deep learning-based surveillance system that detects safety compliance in real time and monitors worker behaviour. The present technique uses the YOLOv8 algorithm to train the deep learning model. If any abnormalities are found on the building site, the project engineer receives an alert immediately. The suggested solution allows constant monitoring of the facility, which improves worker safety. The model is trained using a construction site safety picture dataset obtained from the Roboflow universal datasets. The YOLOv8 model can recognize nine types of safety equipment. Image augmentation techniques are used to increase model accuracy. The suggested algorithm's performance is measured using a variety of measures, including accuracy, recall, the F1- measure, and mean Average accuracy (mAP). This deep learning- based monitoring system runs autonomously and requires no human intervention. The experimental outcomes were analysed using two distinct epochs: 50 and 100. The findings show that the YOLOv8 algorithm has the greatest accuracy of 0.94 when tested across 100 epochs.