Accurate face mask detection, vital for public health, remains challenging in diverse real-world settings due to varied conditions. This research enhances YOLOv8 for face mask detection by integrating an EfficientNet backbone and applying targeted fine-tuning to address these challenges. A Kaggle dataset (YOLO-formatted) was used for model implementation and training on Google Colab. Experimental results confirm our EfficientNet-backed model significantly surpasses baseline YOLOv8 in detection accuracy and robustness across diverse conditions. Specifically, the fine-tuned, EfficientNet-enhanced YOLOv8 shows a notable mAP increase, proving its efficacy for robust, real-time detection, even for minority classes.