This work presents a deep learning-based framework for automatic Personal Protective Equipment (PPE) detection in healthcare environments. The proposed approach is based on a YOLO26n one-stage object detection architecture designed to achieve an effective trade-off between detection accuracy and real-time inference performance. A unified dataset was constructed by merging two publicly available healthcare-oriented datasets, resulting in 5,340 images and 13,308 annotated instances across four PPE categories: Coverall, Gloves, Goggles, and Mask. The model was trained using a transfer learning strategy and evaluated on an independent test set using standard COCO metrics along with Precision, Recall, and F1-score. Performance evaluation indicates that the proposed YOLO26n-based framework can accurately identify PPE items, yielding a mAP@0.5 of 0.944 together with Precision, Recall, and F1-score values of 0.912, 0.944, and 0.928, respectively. Additionally, the model achieves an average inference time of approximately 12.46 ms per image, demonstrating its suitability for real-time applications. Comparison with Faster R-CNN, YOLOv8n, and YOLO11n indicates that YOLO26n achieves the most favorable compromise between detection accuracy and inference efficiency. These results confirm the effectiveness of the proposed approach for real-time PPE monitoring in healthcare environments, where both accuracy and low-latency inference are essential requirements.

Personal Protective Equipment,Object Detection,YOLO,Deep Learning,Healthcare Monitoring,Computer Vision