NurzadaEnu/Personal-Protective-Equipment-Detection-using-YOLOv10: Modeling the impact of dataset size and1 class imbalance on YOLOv10-based PPE2 detection systems3

Personal-Protective-Equipment-Detection-using-YOLOv10 Modeling the impact of dataset size and class imbalance on YOLOv10-based PPE detection systems. ‎Title – Modeling the impact of dataset size and class imbalance on YOLOv10-based PPE detection systems. ‎Description – This paper investigates the influence of dataset size and class imbalance on the performance of YOLOv10 based personal protective equipment (PPE) detection systems. Six YOLOv10 configurations (n, s, m, b, l, x) were tested on domain-specific datasets covering construction, industrial, and medical contexts. A novel mathematical model is introduced to describe the nonlinear relationship between dataset size and detection performance (mAP@50), revealing a saturation threshold beyond which additional data yield diminishing returns (R² = 0.968). The analysis also highlights that the superior accuracy of complex YOLOv10-x models significantly declines under conditions of pronounced class imbalance. These findings underscore the importance of balanced and sufficiently large datasets in optimizing detection accuracy for real-world safety applications. ‎Dataset Information: CHV Dataset SHELK5K SH17 Code Information: In this repository you can find the yolov10 experiment results. ‎ Usage Instructions – Run YOLOv10 models on three domain-specific datasets (CHV, SH17, and SHEL5K) prepared in YOLO format (train/val/test). For each dataset, use the corresponding .yaml configuration file and execute the standard train → val → export workflow to compare mAP, Precision, and Recall across YOLOv10 configurations (n, s, m, b, l, x). ‎Requirements – python >=3.8 and ultralytics>=8.1.0, PyTorch>=1.8.