Composite Artificial Intelligence

9A Novel Transfer Learning-Based Composite AI Model for Skin Disease Classification

R. Karthick Manoj^1* and S. Aasha Nandhini²

¹Department of Electrical and Electronics Engineering, AMET Deemed to be University, Kanathur, India

²Department of Electronics and Communication Engineering, Sri Sivasubramaniya Nadar College of Engineering, Kalavakkam, India

Abstract

Early and accurate diagnosis of skin diseases like melanoma, psoriasis, and eczema is important for successful treatment and management of skin issues. In remote areas, accessibility and efficiency remain as significant issues. To address these issues, this work proposes a novel method for skin disease detection using dermatological images and composite artificial intelligence models based on transfer learning. The advantage of the method lies in its ability to overcome the constraints of small datasets and also aids in early detection. The composite AI model combines the advantages of ResNet50, VGG16, and InceptionV3 architectures through a weighted average approach which aggregates the performance of the individual models. The Model weights are assigned based on performance metrics such as accuracy, precision, and recall evaluated during preliminary testing. This weighting strategy helps to make the composite model more balanced and robust in terms of disease diagnosis. Using Kaggle dataset for training and evaluation, the models are optimized in terms of disease class. At skin disease classification, the composite ...