Industrial sectors face significant challenges in ensuring worker safety due to the presence of heavy machinery and hazardous materials. Current safety protocols often rely on manual processes, which are prone to human error and inefficiency. This study proposes a novel integration of computer vision, deep learning, and machine learning techniques for real-time safety monitoring using live video feeds. Our approach aims to detect and address safety deviations promptly, facilitating proactive measures to prevent accidents and enhance overall safety and operational efficiency. The system’s adaptability allows for customizable detection parameters tailored to specific site requirements, making it suitable for diverse industrial environments and regulatory frameworks. Furthermore, real-time notifications delivered through a data visualization dashboard improve safety oversight, saving time and enhancing efficiency by pre-emptively identifying safety breaches. By integrating the Internet of Things, advanced algorithms like faster R-CNN emerged as the most accurate, achieving the highest precision (87%), recall (84%), F1-score (85.5%), and mAP (80%) compared to other algorithms such as SSD and RetinaNet. Given its excellent accuracy and capability to produce precise detections, even with a computationally demanding two-stage detection approach, faster R-CNN was selected. This system ultimately offers a comprehensive solution with an overall accuracy of 96%, elevating safety standards across industrial operations.