Skin Disease Classification using Transfer Learning with a Modified ResNet-50 Architecture

This paper presents an advanced approach in skin disease classification using a modified ResNet-50 architecture, applied to a specific subset from the ISIC 2019 Dataset focusing on Benign Keratosis, Basal Cell Carcinoma, and Melanoma. The study integrates deep learning and medical imaging to enhance diagnostic accuracy in dermatology. It leverages transfer learning, adapting the well-established ResNet-50 model, and modifies it by incorporating global average pooling and dense layers, thus tailoring it for dermatological use. This method is inspired by significant strides in skin cancer detection using ResNet and innovative transfer learning techniques in skin disease classification. A notable aspect of this research is the evaluation of data augmentation's impact on model generalization and robustness. The findings of this study contribute to the growing field of deep learning in skin disease classification, providing a refined and effective tool for potential applications in dermatology.