Authors : Anil Kumar Reddy Tetali; Tatavarthi Rishi Varun; Guttula Soma Durga Arjun; A. V. N. B. Pavan Kumar; Karri Jaya Naga Sri; T. J. V. Adinarayana

Volume/Issue : Volume 11 - 2026, Issue 3 - March

Google Scholar : https://tinyurl.com/mvfjpezd

Scribd : https://tinyurl.com/3nukbdwr

DOI : https://doi.org/10.38124/ijisrt/26mar1074

Note : A published paper may take 4-5 working days from the publication date to appear in PlumX Metrics, Semantic Scholar, and ResearchGate.

Abstract : Although the automated detection of helmets remains a cornerstone of modern occupational safety and transit safety, the continued prevalence of traumatic head injuries indicates a systemic failure in regulatory compliance. The inherently fragmented nature of manual detection would inevitably collapse under the logistical burden of a large industrial facility or the high-speed traffic flows of urban landscapes. In such scenarios, a paradigm shift from the limitations of human detection to the high-fidelity architectures of autonomous computer vision systems is a necessity, providing a 24-hour state of watchfulness. The efficacy of the YOLO framework, a paradigm-shifting single- stage detector, in the detection of protective headgear in a dynamic environment is the subject of the following investigation. To understand the performance of the YOLO framework, it is essential to understand the architectural divide between multi- stage detection systems and single-stage detection systems. While the traditional two-stage detection systems, such as the Faster R- CNN and Regional Proposal Networks, focus on the accuracy of localization at the cost of considerable computational latency, the single-stage detection systems such as the SSD, RetinaNet, and the YOLO family focus on a unified regression problem. In the weighing of the competing demands of mean Average Precision (mAP) performance and real-time throughput, the choice falls squarely on YOLO as the preeminent choice for safety systems on the edge. The need to react instantaneously to the situation on the road precludes the use of computationally intensive approaches, which would necessarily sacrifice accuracy on the altar of speed. The ability of YOLO to deliver high frame rates without sacrificing accuracy on a catastrophic scale provides the fluidity of action to proactively mitigate safety hazards.

Keywords : Helmet Detection, Intelligent Transportation Systems, Deep Learning, YOLOv8, Object Detection, Traffic Surveillance, Computer Vision.

References :

1. R. Girshick, J. Donahue, T. Darrell, and J. Malik, “Rich feature hierarchies for accurate object detection and semantic segmentation,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Jun. 2014, pp. 580–587.
2. R. Girshick, “Fast R-CNN,” in Proc. IEEE Int. Conf. Comput. Vis. (ICCV), Santiago, Chile, Dec. 2015, pp. 1440–1448.
3. S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards real-time object detection with region proposal networks,” in Proc. Adv. Neural Inf. Process. Syst. (NIPS), Montréal, Can- ada, 2015, pp. 91–99.
4. J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Unified, real-time object detection,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Las Vegas, NV, USA, Jun. 2016, pp. 779–788.
5. J. Redmon and A. Farhadi, “YOLO9000: Better, faster, stronger,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Honolulu, HI, USA, Jul. 2017, pp. 6517–6525.
6. J. Redmon and A. Farhadi, YOLOv3: An incremental improvement, arXiv:1804.02767, 2018.
7. A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, YOLOv4: Optimal speed and accuracy of object detection, arXiv:2004.10934, 2020.
8. Ultralytics, YOLOv5 official documentation, 2020. Available: https://github.com/ultralyt-ics/yolov5.
9. Ultralytics, YOLOv8 official documentation, 2023. Available: https://docs.ultralytics.com/.
10. W. Liu et al., “SSD: Single shot multibox detector,” in Proc. Eur. Conf. Comput. Vis. (ECCV), Amsterdam, Oct. 2016, pp. 21–37.
11. T.-Y. Lin et al., “Focal loss for dense object detection,” in Proc. IEEE Int. Conf. Comput. Vis. (ICCV), Venice, Italy, Oct. 2017, pp. 2980–2988.