Deep Learning-based Multiclass Skin Disease Classification: Comparative Assessment of VGG19, YOLOv3, and MobileNet

The diagnosis of skin diseases by mere inspection can be a tough ask considering the distinctive and complicated nature of skin lesions. This study looks into the application of three advanced Convolutional Neural Network (CNN) models – VGG19, YOLOv3, and MobileNet – for multiclass skin disease classification, to aid diagnosis accuracy and improve accessibility in the field of dermatology. A comparative study of these models was done with the help of a dataset consisting of images of common skin diseases to assess their diagnostic capabilities. The results demonstrate that VGG19 provides the highest accuracy, which is 0. 928. There is a 95% confidence level and the Matthews Correlation Coefficient (MCC) is 0.89 displaying that mobilenet is a powerful tool for feature extraction and class of disease. MobileNet's performance was incredibly strong, but its outperformers were noted for their computational efficiency which makes them suitable for mobile apps. Although YOLOv3, which is mostly intended for object detection, generally functions well with moderately good diagnostic results, this extends the possibility for YOLOv3 application in broad imaging classification tasks. The study showcases the possibility of CNNs in the automated classification of skin disease which can help in reducing the cost by providing more precise and easily-accessible diagnostic solutions. Such tools are not only helpful for the increased accuracy of diagnostics but also for making dermatological services more available, especially for the areas with limited medical expertise.