Advancing Monkeypox Detection with Deep Learning: A High-Accuracy ResNet50-Based Approach

This work aims to create a deep-learning model that will be able to classify Monkeypox and Non-Monkeypox images accurately. Using 3192 equally distributed images as data., trained a ResNet50 pre-trained CNN model to classify the input images between the two classes. The model converged in 50 epochs with an average of 2.,554 training images and 638 validation images., employing several steps such as data augmentation and hyperparameter optimization to detect overfitting. We managed to achieve an average accuracy of 89.97% in our proposed model. The training loss was reduced to 0.0703 from 0.6303., and the validation loss was reduced to 0.1725 from 0.5069., showing good learning and less overfitting. The classification report also indicates an incredible accuracy of both the precision (~90<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>) of the classifier as well as the recall (~90<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>) of the classifier; the F1-Scores give a good sense of measuring an aspect of the sensitivity and specificity of the classifier. The confusion matrix analysis shows a near-equal distribution of misclassifications., highlighting the model's robustness and reliability in distinguishing between the two classes. The significance of this research lies in its potential to assist healthcare professionals in the timely and accurate detection of Monkeypox., thereby facilitating prompt intervention and containment measures. By providing a reliable automated diagnostic tool., this study contributes to advancing Monkeypox detection and underscores the vital role of deep learning in addressing emerging public health challenges.