DETEKSI DAN PENANGANAN PENYAKIT KULIT BERBASIS WEB DENGAN CNN-SVM DAN GEMINI

Abstract

This research develops an image-based classification system to detect eight skin diseases cellulitis, impetigo, athlete’s foot, nail fungus, ringworm, cutaneous larva migrans, chickenpox, and shingles through an interactive web application. The system uses transfer learning with MobileNetV2 pretrained on ImageNet to extract salient visual features such as texture and color patterns from skin images. These features are classified by a Support Vector Machine (SVM) with a linear kernel, generating accurate and efficient predictions. Unlike previous studies that focused solely on model development or provided an interface without supplementary guidance, this system integrates classification and follow-up information. Via a simple and user-friendly interface, users upload a photo of a skin lesion through a browser and immediately receive classification results along with confidence scores. The system also forwards its prediction to the AI Gemini model, which supplies additional details, including disease descriptions, primary symptoms, common treatments, safe self-care guidelines, and advice on when to seek professional care. Performance evaluation shows that the system achieves an accuracy of 0.97, with an average precision of 0.98, an average recall of 0.97, and an average F1-score of 0.97 confirming consistent classification across all disease categories. Overall, this system not only functions as an early diagnosis tool, but also as an educational medium that supports early treatment and decision-making by general medical personnel.