Enhancing workplace safety using YOLOV8: A comparative study

Construction sites are inherently hazardous due to their dynamic and unpredictable environments. These risks are often exacerbated by disorganized management, undisciplined workers, and inadequate safety protocols, leading to a higher probability of accidents that negatively affect both workers and business stakeholders. Traditional surveillance methods—typically involving cameras and manual monitoring—have long been used to mitigate these issues. However, human-monitored systems are prone to errors, delayed response times, and fatigue, limiting their effectiveness. To address these challenges, this study explores the application of Artificial Intelligence (AI), specifically computer vision, to enhance construction site safety monitoring. We utilize the YOLOv8 (You Only Look Once version 8) model, a state-of-the-art object detection algorithm, trained on a relevant construction safety dataset using Google Colab. The model’s performance is then evaluated against human observers. Results show that the AI model achieves an accuracy rate of 89%, compared to 92% for trained human observers and 81% for untrained individuals. These findings suggest that AI-based monitoring systems can provide a viable and effective alternative to traditional human surveillance. Integrating AI in construction site safety management has the potential to improve hazard detection, ensure quicker response times, and significantly reduce the likelihood of injuries or fatalities. This study highlights the promise of AI as a proactive tool for enhancing safety and operational efficiency in the construction industry.