ABSTRACT
Skin diseases are notably usual globally, affecting hundreds of thousands of human beings. Early detection and prognosis of skin conditions is important for timely healing intervention. However, constrained get admission to to dermatologists makes it hard to achieve this goal. Recent improvements in artificial intelligence, specifically deep learning using convolutional neural networks (CNNs), provide a possibility to expand computerized systems for pores and skin disorder prognosis. This paper gives a pores and skin sickness classification framework using CNNs for automated evaluation of dermatological images. A pre-trained Xception CNN version is leveraged for function extraction from pores and skin images. Additional absolutely related layers are introduced and the complete network is trained stop-to-stop on a huge dataset of nineteen, 500 photos spanning 23 skin sickness training. Extensive data augmentation is finished to amplify the diversity of the training facts. Class weights are calculated to handle elegance imbalance. The model achieves brilliant performance with a validation accuracy of ninety seven% and F1 rating of 0.86. A FastAPI internet interface is advanced for actual-time inference by using importing query snap shots. The predicted label, disease summary, commonplace signs and symptoms and remedy length are returned. The machine demonstrates the capability of deep mastering for automatic evaluation of pores and skin lesions. It can help in huge-scale screening and early prognosis of more than one pores and skin diseases. With further research, such AI structures can be efficiently used in medical settings and gain sufferers globally.
