The Classification of Dermatologic Image and Diagnostic System Based on Deep Learning

Skin is the first line of physiological defense of the human body, and cutaneous venereal disease has become a major problem in the medical field. With the development of medical level, the treatment of cutaneous venereal diseases has also been significantly improved, but it still faces great challenges in terms of difficult diagnosis, small coverage, high treatment cost and low efficiency of doctors' diagnosis. To solve the above problems, this paper uses cutting-edge deep learning algorithms, including Inception V3, GoogLeNet, and EfficentNet, to classify skin disease images. Based on the above deep learning algorithms, an online intelligent clinical diagnosis and treatment system is developed to assist doctors in disease diagnosis. The result shows that EfficientNet is the best in the diagnosis of skin disease. EfficientNet is 83.45%, Inception V3 (63.24%), and GoogleNet (75.32%). The result also shows that EfficientNet provides a more efficient model with the same dataset which has the capability to process more complex medical image classifications task.