Skin Sense AI: Detecting Cancerous Skin Diseases using CNN for Multiclass Classification

It is crucial that a timely and precise detection of skin cancer is helpful in maximizing patient prognosis. This paper presents a novel skin cancer detection and classification system capable of classifying seven distinct skin cancer conditions. The system is developed as a web-app with a user-friendly interface and leverages a custom-designed deep learning architecture utilizing convolutional neural networks (CNNs) for multi-class classification. Our model incorporates feature selection techniques to extract details from the image data, allowing for improved differentiation between various skin cancer types. The model is trained on the widely recognized HAM10000 dataset, by meticulously categorizing the images into the seven targeted skin cancer classes. The effectiveness of the proposed system is validated by achieving an impressive micro-averaged accuracy of 92.5%.