Automated Skin Disease Detection System Using the YOLOV8 Deep Learning Model

Abstract

Skin diseases are a significant worldwide health burden, with conditions varying from benign (acne) to potentially lethal (melanoma). Timely and accurate detection plays a critical role in treatment and patient outcomes; unfortunately, traditional methods of detection are heavily reliant on dermatologists, taking time, being subjective and not always available. To overcome these limitations, this research aims to develop an automated real-time skin disease detection system based on the YOLOv8 deep learning approach. The dataset used is a Kaggle dataset of 22 skin diseases. The images were resized, normalised, converted to YOLO format and data augmentation was applied to enhance generalisability. We used precision, recall, F1-score, and mean Average Precision (mAP) for model evaluation. The model's performance with a precision of 90.7%, recall of 89.2%, F1-score of 89.9%, and [email protected] of 0.918 indicated robust performance and detection accuracy. Per-class analysis also confirmed strong performance for diseases that have distinct features, but some misclassifications were noted for similar diseases. Overall, the system shows that YOLOv8 performs well for real-time detection and classification of skin diseases. The system’s speed and accuracy suggest potential for integration into mobile health apps, telehealth services, and decision support systems, though real-world deployment validation remains as future work.