Deciphering Poxvirus Strains: Feature-Driven classification of Monkeypox

This research outlines a method for monkeypox virus classification from images, emphasizing the effective management of augmented datasets to ensure balanced class distribution and provide a robust foundation for model training. The study harnesses the capabilities of advanced pre-trained deep learning models, specifically ResNet50, EfficientNetB3, and DenseNet169, utilizing transfer learning techniques to adapt them to the task at hand. Techniques such as batch normalization and dropout are used to reduce overfitting and improve the generalization capacity of the models. Then, using the optimizer and categorical cross-entropy loss, these models are fine-tuned for monkeypox virus classification. One noteworthy addition is the inclusion of interactive learning, which gives users dynamic control over the training process by enabling them to intervene at predetermined intervals. Performance metrics such as accuracy, confusion matrix are used in the evaluation of the model. In order to guarantee a trustworthy evaluation of the models’ performance, an independent test dataset is utilized. The most impressive accomplishment is the ensemble learning method, which combines the predictions from all three models,ResNet50,EfficientNetB3 and DenseNet169 with a simple averaging procedure. The classification report shows that this all-encompassing methodology consistently produces excellent results, with precision, recall, and F1-score consistently exceeding 0.97. In a time when healthcare challenges are constantly changing, this method shows promise as a useful tool for image-based diagnosis.