A comparative evaluation of CNN architectures for skin disease classification: VGG16 and VGG19

Skin is the largest organ in the human body. However, the skin is very susceptible to infection and can affect a person’s quality of life. Skin is a disease that contributed a lot to global disability in 2013. From this analysis, research is needed regarding the classification of skin diseases. This work divided the classes into 8 sections, including the bacterial ınfections- cellulitis, bacterial ınfections- impetigo, fungal ınfections - athlete’s foot, fungal ınfections - nail-fungus, fungal ınfections - ringworm, parasitic ınfections - cutaneous-larva-migrans, viral skin infections - chickenpox, and viral skin infections - shingles. We implemented two deep (CNN) architectures specifically VGG16 and VGG19 to classify the skin disease images data. The research methodology used is data collection, data exploration, data transform and load, models building, models training, models testing using train data, evaluation of each model, and comparing the models. For VGG16, validation accuracy was obtained at 91.12% and testing accuracy at 89%. Then for VGG19, validation accuracy was obtained at 90.09%, and testing accuracy at 87%. From the results obtained, it can be concluded that VGG16 is superior when compared to VGG19.