Performance Evaluation of EfficientNet Transfer Learning Model for Detecting Face Mask Use with Multi-Class Dataset

Abstract

The spread of COVID-19 Many people are negligent and not alert to the spread of the virus. Many people do not comply with health protocols. The use of personal protective equipment (PPE) to protect physical health, especially the incorrect and inappropriate use of masks. Many people do not comply with the use of masks, which are classified as Personal Protective Equipment (PPE). The use of masks is crucial in reducing the spread of the disease and is essential for maintaining health. To reduce these adverse effects, we use computer technology for face mask detection tools, using the EfficientNet model with a comparison of the B0 and B4 architectures. The results of the comparison between the EfficientNet-B4 model with 50 epochs and the EfficientNet-B0 model with 50 epochs show that the EfficientNet-B4 model has the best results, namely Train Acc (0.9249), Accuracy (0.9523), Train Loss (0.2193), Loss (0.1595), Precision (0.9515), Recall (0.9498), and F1 Score (0.9505). This research uses a dataset as research material and aims to evaluate the Transfer Learning EffienctNet model in detecting the use of face masks properly, correctly, and accurately.