Skin Disease Prediction System

Abstract

Skin cancer is becoming a serious health issue all over the world. If it is detected early, the chances of curing it are very high. However, in many places, it is difficult to find a skin specialist (dermatologist) quickly, and booking an appointment can take weeks or even months. This delay can sometimes be dangerous for the patient. To address this problem, I developed a Skin Disease Prediction System for my final year project. This system is a web-based application that uses Artificial Intelligence to identify skin diseases from images. For the core of the project, I used a Deep Learning model called EfficientNetB3. I chose this specific model because it offers high accuracy while being efficient enough to run on standard computers. I trained the model using the HAM10000 dataset, which contains thousands of examples of common skin lesions. For the application side, I built the backend using FastAPI because it is fast and easy to integrate with Python machine learning libraries. The system allows users to simply upload a photo of a skin lesion, and within seconds, it predicts the type of disease (such as Melanoma or Basal Cell Carcinoma) along with a confidence score. This project aims to serve as a helpful assistant for doctors to speed up their work and as a screening tool for general people to check their skin conditions at home.