Performance Evaluation of Transfer Learning Based Models On Skin Disease Classification

Skin is a vital organ in our body and it even serves as a shield against harmful rays. If skin gets some of the problems, then it would affect the whole body. Major reason for the skin cancer is due to ultraviolet rays, exposing body to chemical and genetical problems. It can spread to the body's inner cells and cause DNA damage. Diagnosing skin cancer at earlier stage becomes more important in order to decrease the malignancy of the disease. Deep learning technology can solve the complexity and helps us to diagnose skin cancer. Deep learning is a part of machine learning where it can work like human brain and can work with its past experience of solving complex problems. Convolutional neural networks can classify skin lesions with much higher accuracy so in this experiment, deep learning model is used. Convolutional neural networks, which may have up to 30 layers, are designed exclusively for image analysis. Categorization, image processing, and segmentation are some of the applications. In this study, transfer learning models including VGG16, VGG19, Inception V3, Convolutional neural networks, and ResNet were used. Benefits of CNN is the ability to develop internal representation of 2D image and it can learn more than SVM, KNN etc which becomes key point in using CNN.