Implementasi Metode Convolutional Neural Networks untuk Klasifikasi Penyakit Kulit dengan Perbandingan Arsitektur VGG-19 dan ResNet-101

Abstract

The skin is the body's outermost organ that plays an important role in protecting internal organs from damage and attack by disease-causing pathogens such as bacteria, microbes and viruses. Skin diseases have many types, forms and causes, from non-contagious to contagious and even chronic. Convolutional Neural Networks (CNN) is needed as a classification of an image to detect, because CNN has the ability to classify and get the most significant results in image recognition, which is intended for image data as a solution in the classification of facial skin diseases. In this study, facial skin diseases will be classified using the CNN method with two architectural models VGG19 and ResNet101 which are included in deep learning. Based on the research that has been done, to classify the image of skin diseases on the face using CNN with two architectural models VGG19 and ResNet101 obtained by adding a dropout layer with a value of 0.5 and the number of dense layers as many as 32 by using the RELU activation function and using several appropriate parameters based on the test scenario, namely with an input shape of 224 x 224 pixels, epochs 50, batch size 20, optimizer Adam, learning rate 0.001 and with data scenarios 80:10:10. The accuracy obtained from the two architectural models used is 93.33% for VGG19 and 76.32% for ResNet101.