Deep Learning in Dermatology: Convolutional Neural Network-Based Classification of Skin Diseases and Cancer

Skin diseases are a widespread condition. Clinically monitoring dermatological disorders and classifying them is a complex process. Because of the utter complexity of individual skin tones and the evident vicinity effect of disorders, diagnosing the specific disorder is difficult. This research paper aims to achieve a high percentage of accuracy in skin disease image classification using deep learning models. The dataset chosen is a skin disease and skin cancer dataset consisting of 57 different classes. The reason behind using deep learning is that it has outperformed human doctors in diagnosing skin diseases on several occasions. A custom model has been developed for research which uses convolutional neural network achieving an accuracy of over 96.64%. This study also compares the performance of this custom model with other pre-trained models which are deemed to be highly accurate for the work of image classification and are frequently used in developing high-end models for a biomedical use.