Enhancing Public Safety through Face Mask Detection using Xception Deep Learning Architecture

Recent developments in controlling the ongoing epidemic include the widespread use of face masks. This creates an urgent need for precise and fast analytical systems for face mask detection to ensure public safety from viral infectious diseases. This paper proposes an optimal and innovative mask detection method based on the Xception deep learning model to automatically identify faces, whether masked or not. The performance was tested utilizing an effective dataset of 11,792 pictures, segregated into two categories: With Mask and Without Mask. The model was trained with the Adam optimizer and binary cross-entropy loss function to achieve the desired performance. As a result, the model trained on the training dataset demonstrated a high accuracy of 98.79% on the test dataset. Additionally, the results revealed high precision, recall, and F1 scores, all of which were above 0.99 for the two categories. This is due to a high level of precision and recall at multiple average data points but low errors in classification. A confusion matrix and a classification report were developed to demonstrate the comprehensive models' performance. Therefore, the developed method will deliver a reliable system for real-world face mask detection with strong performance characteristics. This study adds value by offering a more precise, fast, and innovative model for a mask detection system that answers the needs for overall public security.