Paper Details
Abstract
Early diagnosis of skin diseases plays a critical role in improving treatment effectiveness and reducing societal burden. However, traditional deep learning models often provide only classification results without offering explanations or insights into decision-making. This study proposes integrating the Recipro-CAM explainability technique with the ResNet-50 model to achieve high classification accuracy and enhance interpretability. The research evaluated a dataset comprising 10,000 images of four common skin conditions (Monkeypox, Measles, Chickenpox, and Acne). The proposed method achieved over 98.80% accuracy, precision, recall, and F1-score, while the interpretability evaluation based on Intersection over Union (IoU) exceeded 0.46. These results demonstrate the potential of explainable artificial intelligence (XAI) to clarify deep learning predictions and pave the way for further research to improve diagnostic systems' accuracy and transparency.