This work introduced the Hardhat10K dataset with over 10,000 images for hardhat-wearing detection. Images containing long-distance, occluded, dense, and low-light objects were collected to enhance the model's robustness. Furthermore, images from various weather conditions and periods were added to improve the model's generalization ability. Finally, 300 background images were supplemented to enhance the model's accuracy. The dataset was annotated with YOLO format and categorized into six classes: "hardhat", "head_with_hardhat", "person_with_hardhat", "head", "person_no_hardhat", and "face".
You can download the hardhat10k dataset through the https://drive.google.com/file/d/1DRnGkcRxM2mu4Xld3FV18Cc2ymS-CJk4/view?usp=drive_link.
[1] Larxel, “Safety Helmet Detection,” https://www.kaggle.com/andrewmvd/hard-hat-detection, accessed on May 3, 2024.
[2] M.-E. Otgonbold, M. Gochoo, F. Alnajjar, L. Ali, T.-H. Tan, J.-W. Hsieh, and P.-Y. Chen, “SHEL5K: An Extended Dataset and Benchmarking for Safety Helmet Detection,” Sensors, vol. 22, no. 6, pp.2315-2337, 2022.
[3] JUNWIDE, “SafetyHelmetWearing,” https://www.kaggle.com/datasets/junwide/safetyhelmetwearing, accessed on May 3, 2024.
[4] L. B. Xie, “Hardhat,” https://doi.org/10.7910/DVN/7CBGOS, accessed on May 3, 2024.
[5] L. B. Xie, “Hard Hat Workers Dataset,” https://public.roboflow.com/object-detection/hard-hat-workers, September 2022.
ATTN: This dataset is only for academic usage.
Wanbo Luo, Ahmad Ihsan Mohd Yassin, Khairul Khaizi Mohd Shariff, Rajeswari Raju
Universiti Teknologi MARA