Personal Protective Equipment Detection Method Based on Knowledge Distillation

Purpose: Automatic detection of Personal Protective Equipment (PPE) is critical for ensuring safety in construction sites. However, current methods struggle with two major challenges. One is the limited diversity of training data, which restricts robustness in adverse environments. The other is the accuracy–efficiency trade-off inherent in lightweight models. Methods: We propose a detection framework called Distillation Alignment YOLO (DA-YOLO) for PPE detection. Our method enhances data diversity through controllable augmentation of adverse conditions and introduces a teacher–student distillation framework with consistency constraints across predictions and features. This joint constraint enables the student model to achieve strong generalization while maintaining low computational cost. Results: Experimental results on public PPE dataset show that DA-YOLO achieves the best performance under the same parameter size with mAP50 reaching 92.81%, representing an improvement of 5.5% over YOLOv13. Conclusion: By integrating controllable augmentation and a dual-consistency distillation strategy, our method mitigates the critical issues in smart construction site.