Vision Transformer Framework Approach For Melanoma Skin Disease Identification

Abstract

Abstract In the past few decades, skin diseases have been a hazardous issue because of more sophisticated and high-cost treatments. Identifying skin disease is still a challenging task for dermatologists. In reference to severe diseases like Melanoma, therapy in the initial stages is very important and effective to avoid skin cancer. This paper proposes an effective approach by using Vision Transformers (ViT) to detect Melanoma, which gives the accuracy of 99% on the test images. Authors considered the dataset, which is publicly available on Kaggle that comprises 1000 images and did the comprehensive study to get better results using ViT. The obtained results are compared with other state-of-the-art algorithms (VGG-19 and Inception-V3) to analyze the distinction between the proposed approach and other Convolutional Neural Network (CNN) Models.