A Survey on Skin Disease Classification using Pre-trained Deep Learning Algorithms

Abstract

Skin diseases are most common in people’s everyday life. Every year, millions of people are affected by many types of skin disorders which induce serious impact on their health. Since the human analysis of such diseases takes some time and effort and recent algorithms are only used for analyzing single class of skin diseases, there is a need for a more high-level of computer-aided expertise in the analysis of multi-class skin diseases. In recent methods, Transfer Learning (TL) with deep learning algorithms is used to train Deep Neural Network (DNN) via pre-learned models. There are different deep learning frameworks designed for classifying various kinds of skin diseases using different skin image datasets. In this article, a survey on different deep learning algorithms for classifying various skin diseases or melanoma diseases is presented. Also, a comparative analysis is presented that includes the merits and demerits of the surveyed algorithms for skin diseases classification in a tabular form. Finally, the future directions are suggested based on the limitations observed in those algorithms for enhancing the efficiency of skin diseases classification.