Heterogeneous Multi-Model Ensemble for PPE Detection in Construction Environments
DOI:
https://doi.org/10.4108/eetiot.9971Keywords:
PPE, Detection, Ensemble, Multi-Model, Construction EnvironmentsAbstract
The construction industry remains one of the most hazardous work environments, with a fatality rate of 25.6 per 100,000 workers, significantly exceeding the average across all industries. PPE compliance is crucial for worker safety, yet monitoring adherence remains challenging in dynamic construction environments. This paper presents an automated PPE detection system utilizing an ensemble deep learning models to enhance workplace safety monitoring. Our approach combines three advanced architectures with Yolov11, RTDETRv2, and HyperYolo. The individual model predictions are integrated using WBF to improve detection robustness. We evaluate our system on a comprehensive dataset of 4,135 professionally annotated images encompassing critical PPE categories including hard hats, safety vests, protective gloves, and safety boots, along with their corresponding absence classes. The proposed ensemble achieves superior performance with a precision of 0.765, recall of 0.735, mAP@50 of 0.760, and mAP@50:95 of 0.440, outperforming individual models across all evaluation metrics. The results demonstrate the effectiveness of multi-model fusion for automated PPE detection. This research contributes to the advancement of intelligent safety systems that can significantly reduce workplace injuries and fatalities through automated PPE compliance verification.
Downloads
References
[1] Bureau of Labor Statistics, U.S. Department of Labor. (2023). Construction deaths due to falls, slips, and trips increased 5.9 percent in 2021. Available at: https://www.bls.gov/opub/ted/2023/construction-deaths-due-to-falls-slips-and-tripsincreased-5-9-percent-in-2021.htm (accessed March 24, 2025).
[2] Abukhashabah, E., Summan, A., & Balkhyour, M. (2020). Occupational accidents and injuries in construction industry in Jeddah city. Saudi Journal of Biological Sciences, 27(8), 1993–1998.
[3] Kisner, S. M., & Fosbroke, D. E. (1994). Injury hazards in the construction industry. Journal of Occupational and Environmental Medicine, 36(2), 137–143.
[4] Kursunoglu, N., Onder, S., & Onder, M. (2022). The evaluation of personal protective equipment usage habit of mining employees using structural equation modeling. Safety and Health at Work, 13(2), 180–186.
[5] Khoshakhlagh, A. H., Malakoutikhah, M., Park, J., Kodnoueieh, M. D., Boroujeni, Z. R., Bahrami, M., & Ramezani, F. (2024). Assessing personal protective equipment usage and its correlation with knowledge, attitudes, performance, and safety culture among workers in SMEs. BMC Public Health, 24(1), 1987.
[6] Alemu, A. A., Yitayew, M., Azazeh, A., & Kebede, S. (2020). Utilization of personal protective equipment and associated factors among building construction workers in Addis Ababa, Ethiopia. BMC Public Health, 20, 1–7.
[7] Lavanya, G., & Pande, S. D. (2024). Enhancing real-time object detection with YOLO algorithm. EAI Endorsed Transactions on Internet of Things, 10.
[8] Ludwika, A. S., & Rifai, A. P. (2024). Deep learning for detection of proper utilization and adequacy of personal protective equipment in manufacturing teaching laboratories. Safety, 10(1), 26.
[9] Wang, Z., Wu, Y., Yang, L., Thirunavukarasu, A., Evison, C., & Zhao, Y. (2021). Fast Personal Protective Equipment Detection for Real Construction Sites Using Deep Learning Approaches. Sensors, 21(10), 3478.
[10] Do, M. T., Kim, T. D., Ha, M. H., Chen, O. T. C., Nguyen, D. C., & Tran, A. L. Q. (2023, December). An Effective Method for Detecting Personal Protective Equipment at Real Construction Sites Using the Improved YOLOv5s with SIoU Loss Function. In 2023 RIVF International Conference on Computing and Communication Technologies (RIVF) (pp. 430–434). IEEE.
[11] Ke, X., Chen, W., & Guo, W. (2022). 100+ FPS detector of personal protective equipment for worker safety: A deep learning approach for green edge computing. Peerto-Peer Networking and Applications, 15(2), 950–972.
[12] Ahmed, M. I. B., Saraireh, L., Rahman, A., Al-Qarawi, S., Mhran, A., Al-Jalaoud, J., Al-Mudaifer, D., Al-Haidar, F., AlKhulaifi, D., Youldash, M., & Gollapalli, M. (2023). Personal Protective Equipment Detection: A Deep-Learning-Based Sustainable Approach. Sustainability, 15(18), 13990.
[13] Jocher, G., Qiu, J., & Chaurasia, A. (2023). Ultralytics YOLO (Version 8.0.0) [Computer software]. https://github.com/ultralytics/ultralytics
[14] Khanam, R., & Hussain, M. (2024). YOLOv11: An overview of the key architectural enhancements. arXiv preprint arXiv:2410.17725.
[15] Lv, W., Zhao, Y., Chang, Q., Huang, K., Wang, G., & Liu, Y. (2024). RT-DETRv2: Improved baseline with bagof-freebies for real-time detection transformer. arXiv preprint arXiv:2407.17140.
[16] Zhao, Y., Lv, W., Xu, S., Wei, J., Wang, G., Dang, Q., ... & Chen, J. (2024). DETRs beat YOLOs on real-time object detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 16965–16974).
[17] Feng, Y., Huang, J., Du, S., Ying, S., Yong, J. H., Li, Y., ... & Gao, Y. (2024). Hyper-YOLO: When visual object detection meets hypergraph computation. IEEE Transactions on Pattern Analysis and Machine Intelligence.
[18] Nguyen, N. T., Tran, Q., Dao, C. H., Nguyen, D. A., & Tran, D. H. (2024). Automatic detection of personal protective equipment in construction sites using metaheuristic optimized YOLOv5. Arabian Journal for Science and Engineering, 49(10), 13519–13537.
[19] Huang, M. L., & Cheng, Y. (2025). Dataset of personal protective equipment. Data in Brief, 111988.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 Thi-Nguyen Nguyen, Quang-Anh Nguyen-Duc, Dinh-Thai Kim, Duy Tung Doan, Minh-Duc Pham, Duc Cuong Van, Minh-Anh Nguyen
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
This is an open-access article distributed under the terms of the Creative Commons Attribution CC BY 4.0 license, which permits unlimited use, distribution, and reproduction in any medium so long as the original work is properly cited.
