Enhanced Dermatological Diagnosis: Autoimmune and Non-Autoimmune Skin Disease Classification Using MobileNet, ResNet, and YOLOv11

Autoimmune diseases arise when the immune system mistakenly attacks the body's healthy cells, causing a range of symptoms that can greatly affect a patient's quality of life. In Indonesia, these conditions present a significant public health concern. According to research by Kemenkes in 2024, autoimmune lupus affects approximately 0.5% of the population, impacting over 1.3 million individuals. This study proposes a classification and detection model utilizing Convolutional Neural Networks (CNN) with transfer learning, incorporating MobileNetV2, MobileNetV3Small, MobileNetV3Large, ResNet50, ResNet101, ResNet152, and YOLOv11 architectures. The model's performance is assessed using a confusion matrix, evaluating precision, recall, and F1-score, while computational efficiency is analyzed using a GPU T4. Experimental results demonstrate that ResNet152 achieved the highest accuracy at 92%, whereas YOLOv11 recorded the lowest at 46%. These findings emphasize the crucial role of selecting an optimal CNN architecture to enhance the accuracy of autoimmune and non-autoimmune skin disease classification and detection.