In underground mining sites, detecting coal miners is essential to protecting their safety and welfare. Nonetheless, there are many difficulties in this task. Significant obstacles include low visibility, complicated backgrounds, erratic environmental conditions, and a lack of annotated data. Human considerations and real-time processing requirements add to the task’s complexity. To overcome these obstacles, multidisciplinary approaches utilizing cutting-edge sensor technologies, machine learning techniques, and domain-specific expertise are required. Various efforts have been made to detect the mine workers but suffer from several challenges like: Absence of real-time detection system, low detection accuracy. In this study, we have utilized on comprehensive dataset called DSLF+ to fine tune two cutting-edge detection model—DETR and YOLOv7 to accurately detect the coal miners. Intensive experiments have been conducted to show the effectiveness of the proposed method. YOLOv7 has shown a detection accuracy of 90% and DETR model has shown a maximum recall of 90%.