Multiple Nail-Disease Classification Based on Machine Vision Using Transfer Learning Approach

Human nails can be an early indicator of severe diseases. This technique is very common in the medical sector. However, the only shortcoming is the low human-eye capability to analyze vast colors and detect slight differences. Therefore, we have combined image processing techniques and deep learning algorithms to generalize models which can provide a high accuracy rate and speed up the process of detecting diseases. In our paper, we have chosen three types of nail diseases which are- Bluish Fingernails, Red Puffy Nails, and Yellow Fungal Nails. These are the most common nail abnormalities indicating medical severity. We used three CNN (Convolutional Neural Network) models, then compared the training and testing rate to find out the most efficient model. The models are MobileNetV2, VGG16, and VGG19 and the achieved accuracy rates were 92%, 72%, and 89% respectively. These accuracy rates were acquired from testing on previously unseen images.