AI-Based Skin Disease Detection and Classification Using Deep Learning

Abstract

Skin diseases are common health problems that affect people across different age groups and geographical regions. Early and accurate identification is important because delayed diagnosis or inappropriate treatment may lead to complications. This paper presents an AI-based Skin Disease Detection and Classification System that analyzes digital skin images and provides preliminary information about possible skin conditions. The proposed system uses image preprocessing operations such as resizing, pixel normalization, noise handling, and image standardization to improve input consistency. EfficientNetB0 is used as the primary deep learning model with transfer learning to extract meaningful visual features and classify predefined skin disease categories. The predicted disease and confidence score are displayed through an Android-based application. The application is developed using Android Studio, Java, and XML, while the model is trained using Python, TensorFlow, and Keras and deployed using TensorFlow Lite. A product recommendation module maps the detected condition to a predefined recommendation database containing product category, ingredients, intended skin concern, usage information, and safety precautions. The proposed system is intended for preliminary screening and skin-disease awareness and is not a replacement for professional medical diagnosis. Keywords: Convolutional Neural Networks (CNN), Deep Learning, Skin Disease Detection, EfficientNetB0, HAM10000, Transfer Learning, Image Preprocessing, TensorFlow Lite, Android Application