Human Nail Disease Diagnosis Systemx using Deep Learning

This research paper aims to introduce a deep learning-based diagnosis for human nail diseases through an improved MobileNet computer vision model with new convolutional layers. The paper is aimed at categorizing 17 nail diseases: Alopecia Areata; Beau's lines; and Yellow Nails, among others. For better generalizations, the data is augmented using techniques such as rotation, zoom and flips; Further, the image preprocessing done in the pipeline provides precise classification. The model is trained using categorical cross entropy and early stopping methodology and model checkpointing to avoid over fitting of the model. The suggested system was capable to yield a significant rate of classification of differential diagnose of various nail diseases. The results shows the success of the classes; the precision, recall, and F1-score for each class are calculated. The paper concludes that the model could be used as an aid to identification of diseases in clinical set ups with the help of automated image processing.