Deep Learning-Based Skin Disease Detection: Comparative Performance Analysis of YOLOv8 and YOLOv11 Models

Recent advances in the field of deep learning have revealed significant potential for automating the detection of skin diseases by overcoming the limitations of traditional diagnostic methods in terms of speed and accessibility. This study presents a comparative analysis aimed at evaluating the performance of the state-of-the-art object detection models YOLOv8 and YOLOv11 in detecting ten different skin disease classes using a publicly available dataset. The dataset consists of a total of 2,486 images, divided into training, validation, and test subsets, and both models were trained under identical conditions for 100 epochs on a Tesla T4 GPU. To assess model effectiveness, performance metrics including precision, recall, F1-score, and mean Average Precision ([email protected]) were analyzed. The results indicate that, compared to YOLOv8, which achieved 84% precision and 91% recall, YOLOv11 attained 85% precision and 92% recall, demonstrating a limited improvement in both precision and recall metrics. Class-wise analyses revealed high detection performance for conditions such as Kurap (fungal infection), whereas relatively lower performance was observed for classes such as Bisul and Urticaria. Furthermore, YOLOv11 achieves this improvement while reducing the number of parameters by approximately 22%, thereby offering an advantage in terms of computational efficiency. These findings demonstrate that YOLOv11 provides a computationally efficient approach and indicate its potential use as a decision-support tool for dermatologists in resource-constrained environments.