A Deep Learning Approach for Real-Time Defect classification in Skin disease

Abstract

In medical analysis, generally, skin diseases are the one-fourth leading cause of nontoxic disease problem in modern years. This study proposed a high-tech system of classifying skin disease using deep learning-based Visual Geometry Group Network (VGGNet) and CONV architecture that extended with limited changes. The system of the VGGNet model to be improved accuracy and that can work on complex computational devices. The proposed model parameters in the CONV layers and improve preparation time for precise predictions. The proposed model is used MNIST HAM10000 data of images which has 10,015 images and released by ISIC archive and the proposed model has outperformed other methods with more than 92% accuracy. The proposed image processing model performed on the inputs of a dermoscopic image to categorize skin disease using a pre-trained convolutional neural network. Therefore, an automatic approach was applied for this classification task using a deep learning model Fine-Tuned VGG-CNN, to increase the classification performance of CNN in the modeling process.