HYBRID DEEP LEARNING MODELS FOR MULTI-CLASS SKIN DISEASE CLASSIFICATION

Abstract

The aim of this study is to test deep-learning models that can identify different types of skin diseases. This study examines illnesses caused by tiny living things, like germs and parasites. The information used in this study came from Kaggle and included pictures of eight types of skin diseases. These pictures were changed by resizing them, changing their colors, and adding different types to make them look better. The information was split into 80% for training and 20% for testing. Five models—CNN, SVM, ResNet50, VGG16, and Mobile-Net—were created and tested using common measures like precision, accuracy, recall, and F1-score. Transfer learning was used with pre-trained deep learning models (ResNet50, VGG16, and Mobile-Net) to make them perform better on this specific dataset. The test showed that VGG16 and Mobile-Net performed much better than the other models, with both achieving 98% accuracy. VGG16's deeper design helped it find details accurately, while Mobile-Net's simpler design kept it fast without losing accuracy. ResNet-50 achieved 96% accuracy, showing that it is good at handling complex patterns. CNN and SVM performed well, achieving accuracy rates of 94% and 93%..