An Investigation of YOLO-Based Helmet Detection for Workers on GPU and CPU Platforms
Article Sidebar
Main Article Content
Abstract
This research focused on enhancing workplace safety by detecting workers’ helmet usage through the integration of artificial intelligence with closed-circuit television systems (CCTV). The detection process was based on object detection techniques using the YOLO (You Only Look Once) model family. This study evaluates the performance of the latest versions—YOLOv10 and YOLOv11—on both CPU and GPU platforms, considering both detection accuracy and inference speed. Among the tested model variants, a maximum processing speed of 454.52 frames per second was achieved on GPU and 34.01 frames per second on CPU, demonstrating their suitability for real-time CCTV-based applications. In terms of accuracy, YOLOv11n achieved a mean Average Precision (mAP@50) of 98.3% for helmet detection and 96.6% for head detection. The model also demonstrated strong classification performance, with precision and recall of 85.76% and 96.98% for the head class, and 93.02% and 96.56% for the helmet class, respectively. These results confirmed the model’s effectiveness and balanced performance in distinguishing between helmeted and non-helmeted workers, supporting its potential for practical and cost-effective deployment in real-time safety monitoring systems.
Downloads
Article Details
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Journal of TCI is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) licence, unless otherwise stated. Please read our Policies page for more information...
References
CBC News, “Surveyor killed in Quebec road construction accident,” http://www.cbc.ca/news/canada/story/2011/09/02/worker-death-chateauguay.html, accessed Aug. 3, 2013.
CBC News, “3 most dangerous job sectors in Canada,” http://www.cbc.ca/news/canada/story/2012/04/25/f-dangerous-jobs.html, accessed Aug. 3, 2013.
A. Voulodimos, N. Doulamis, A. Doulamis, and E. Protopapadakis, “Deep learning for computer vision: A brief review,” Computational Intelligence and Neuroscience, vol. 2018, no. 1, 2018, doi: 10.1155/2018/7068349.
M. W. Park, N. Elsafty, and Z. Zhu, “Hardhat-wearing detection for enhancing on-site safety of construction workers,” J. Constr. Eng. Manag., vol. 141, no. 9, p. 04015024, 2015, doi: 10.1061/(ASCE)CO.1943-7862.0001005.
E. Wanlin, Z. Yang, and J. Yu, “Detection and recognition of personal protection equipment wearing based on an improved YOLOv5 algorithm,” in Proc. 2023 4th Int. Symp. Comput. Eng. Intell. Commun. (ISCEIC), Aug. 2023, pp. 229–232, doi: 10.1109/ISCEIC57870.2023.10291814.
S. Chen et al., “Detection of safety helmet wearing based on improved faster R-CNN,” in Proc. 2020 Int. Joint Conf. Neural Networks (IJCNN), Jul. 2020, pp. 1–7, doi: 10.1109/IJCNN48605.2020.9206847.
J. Hu, X. Gao, H. Wu, and S. Gao, “Detection of workers without helmets in videos based on YOLO V3,” in Proc. 2019 12th Int. Congr. Image Signal Process., Biomed. Eng. Inform. (CISP-BMEI), Oct. 2019, pp. 1–4, doi: 10.1109/CISP-BMEI48228.2019.8965602.
M. Tan, R. Pang, and Q. V. Le, “EfficientDet: Scalable and efficient object detection,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., 2020, pp. 10781–10790, doi: 10.1109/CVPR42600.2020.01079.
E. Rohadi et al., “Internet of Things: CCTV monitoring by using Raspberry Pi,” in Proc. 2018 Int. Conf. Appl. Sci. Technol. (iCAST), Oct. 2018, pp. 454–457, doi: 10.1109/iCAST.2018.8751583.
A. Kurniawan, A. Ramadlan, and E. M. Yuniarno, “Speed monitoring for multiple vehicle using closed circuit television (CCTV) camera,” in Proc. 2018 Int. Conf. Comput. Eng. Netw. Intell. Multimedia (CENIM), Nov. 2018, pp. 88–93, doi: 10.1109/CENIM.2018.8751039.
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., Jun. 2016, pp. 779–788, doi: 10.1109/CVPR.2016.91.
N. Jegham, C. Y. Koh, M. Abdelatti, and A. Hendawi, “Evaluating the evolution of YOLO (You Only Look Once) models: A comprehensive benchmark study of YOLO11 and its predecessors,” arXiv preprint arXiv:2411.00201, 2024.
S. Visa, B. Ramsay, A. L. Ralescu, and E. Van Der Knaap, “Confusion matrix-based feature selection,” Maics, vol. 710, no. 1, pp. 120–127, 2011.
A. Subasi, Practical Machine Learning for Data Analysis Using Python. Academic Press, 2020.
Steadman, P., Jenkins, P., Rathore, R. S., & Hewage, C. (2024, July). Challenges in Implementing Artificial Intelligence on the Raspberry Pi 4, 5 and 5 with AI HAT. In The International Conference on Computing, Communication, Cybersecurity & AI (pp. 147-157). Cham: Springer Nature Switzerland.
Roboflow, “Hard Hat Workers Dataset,” Accessed: Dec. 24, 2024. [Online]. Available: https://public.roboflow.ai/object-detection/hard-hat-workers
E. Bisong, “Google Colaboratory,” in Building Machine Learning and Deep Learning Models on Google Cloud Platform, 1st ed. Berkeley, CA, USA: Apress, 2019, pp. 59–64, doi: 10.1007/978-1-4842-4470-8_7.
NVIDIA, “GPU positioning for virtualized compute and graphics workloads,” Accessed: Dec. 24, 2024. [Online]. Available: https://images.nvidia.com/data-center/technical-brief-gpu-positioning-virtualized-compute-and-graphics-workloads.pdf
Intel, “Measuring processor power: TDP vs. ACP whitepaper,” Accessed: Dec. 24, 2024. [Online]. Available: https://www.intel.com/content/dam/doc/white-paper/resources-xeon-measuring-processor-power-paper.pdf
AMD, “5th Gen AMD EPYC processor architecture whitepaper,” Accessed: Dec. 24, 2024. [Online]. Available: https://www.amd.com/content/dam/amd/en/documents/epyc-business-docs/white-papers/5th-gen-amd-epyc-processor-architecture-white-paper.pdf
Ultralytics. (2024, September 30). YOLOv11 models. Ultralytics Documentation. Retrieved May 19, 2025, from https://docs.ultralytics.com/models/yolo11/
