Approach towards the Automated System for Understanding Facial Skin Diseases

Abstract

Now a days the facial skin problems can be identified and treated with the significant methods of computer-based technology and artificial intelligence. These skin diseases not only depress the infected person and cause psychological depression, but they may also lead to cause skin cancer. As the visual resolution is not so effective in skin disease images, medical experts with latest technical instruments are required for diagnosis of various types of diseases. These skin disease problems can be diagnosed automatically with the help of these computer aided system. In this paper, we suggested a Deep Convolutional Neural Network (CNN) which is an automated method for facial skin disease classification. In biomedical imaging-based decision making the importance of computer assisted diagnosis in deep learning is identified. ResNETs can be utilized for eliminating deep networks vanishing gradient problems. Various kinds of results may be obtained by the ResNET architectures having various activation functions, batch sizes, tested images and training stages. The activation functions ReLU and SELUare analyzed with the implementation of four network models are observed and by using same data sets image classification is done by residual learning. Highest accuracy is obtained by using ResNET with SELU instead of using residual block may lead to best accuracy rate of 97.01% for classification of various disease images.