Perbandingan Akurasi Model ResNet50 dan VGG16 dalam Mengklasifikasi Penyakit Cacar Menggunakan Metode Convolutional Neural Network

Abstract

Pox disease is an infectious disease that attacks the immune system. This disease is often considered mild because it is usually harmless and does not cause death. With the lack of counseling and treatment of this disease, lacks an understanding of pox disease, the types and the treatment . Some of the types of pox disease are Monkeypox, Chickenpox, and Cowpox. Cowpox can cause complications such as keratitis and corneal melting due to persistent erosion. To avoid complications of pox disease arising, CNN algorithm is considered capable of classifying the type of pox disease. Therefore, this study was conducted using 2 CNN methods used, namely: ResNet50 and VGG16. Because the distinguishing features of pox disease classification involve a mixture of color, shape and texture, CNN algorithm approaches including ResNet50 and VGG16 were tried. The ResNet50 model showed quite good results with an average accuracy of 76% but VGG16 had better accuracy with a value of 93%. The ResNet50 model showed good results with an average accuracy of 76% but VGG16 had better accuracy with a value of 93%. Using further evaluation, VGG16 has superior values with good precision, recall, and F1-score  for each class. This proves that VGG16 is a superior model for classifying pox disease types. The ResNet50 model is good at identifying three classes, namely, Chickenpox, Cowpox, and Healthy compared to the Monkeypox class because that class has 25% recall value and 40% F1-score  value. Similarly, the VGG16 model where the monkeypox class still has a recall value of 76%. This shows the potential of using artificial intelligence technology to classify smallpox disease.