EfficientNetB3 for Rapid Diagnosis of Monkeypox: A Transfer Learning Approach for Enhanced Medical Image Classification

The rapid emergence of Monkey Pox as a global health concern demands an efficient and accurate diagnostic approach to support healthcare professionals. This study presents a robust solution using the EfficientNetB3 model, trained with transfer learning, to classify "Monkey Pox" and "Non-monkeypox" conditions from medical images. With an overall accuracy of 87%, our model demonstrates high precision and recall, ensuring reliable detection and differentiation between these conditions. Training over 25 epochs revealed consistent improvement in performance, with training accuracy reaching 93% and validation accuracy peaking at 88%, indicating strong generalization capabilities with minimal overfitting. The model achieved impressive metrics for both classes, including F1-scores of 0.9048 for "Monkey Pox" and 0.9250 for "non-monkeypox," supported by a detailed confusion matrix analysis that shows only two misclassifications. These results highlight the model's ability to effectively identify positive cases while maintaining low error rates. Our findings suggest that the EfficientNetB3 model can serve as a valuable tool in supporting the early detection and accurate classification of Monkey Pox, offering a promising aid for healthcare providers.