Harnessing EfficientNetB5 for Accurate Skin Disease Classification: A Deep Learning Approach

Skin diseases are best treated when detected at a very early stage and deep learning can be used to help in this identification process. In this work, an EfficientNet deep learning model is introduced to classify 10 types of skin diseases, such as Melanoma, Basal Cell Carcinoma, and Psoriasis. The proposed model has an overall accuracy of 87%, which shows high precision and recall for conditions such as Melanoma and Melanocytic Nevi (NV). Analysis of performance through a classification report and confusion matrix shows that although the model is accurate, there is still a problem in differentiating between Psoriasis & Lichen Planus and Seborrheic Keratoses. The results show that EfficientNet can be effectively applied for dermatological purposes and provide an effective approach to automated skin disease diagnosis. Some potential future work could enhance the model's capacity to distinguish between visually similar diseases and its applicability in clinical practice.