Automated Skin Disease Classification Using Fine-Tuned MobileNetV2 Implemented with TensorFlow and Keras

Skin diseases are among the most widespread and visually diverse medical conditions worldwide, affecting millions of people annually. Accurate diagnosis of dermatological disorders often requires access to experienced dermatologists and specialized imaging tools. However, such resources are limited in many regions, leading to delayed or incorrect diagnosis. This paper presents a deep learning–based automated framework for skin disease classification using a fine-tuned MobileNetV2 model implemented with TensorFlow and Keras. The dataset, compiled from five publicly available Kaggle sources, contains over 56,000 images covering 30 representative disease classes out of more than 2,500 known dermatological conditions. The model achieved a validation accuracy of 29.36% under challenging inter-class visual similarity and dataset imbalance.A bilingual Tkinter-based GUI was developed to provide interactive classification capability in English and Persian, offering an accessible interface for both researchers and general users. Two additional API buttons—one for ChatGPT medical dialogue and another for Drug Store integration—were designed as placeholders for future extensions.The study demonstrates that lightweight convolutional architectures like MobileNetV2 can form the basis for resource-efficient medical image classification systems, offering potential for affordable, scalable AI-driven dermatological assistance.