Dilated Convolutional Neural Network for Skin Cancer Classification Based on Image Data

Skin cancer is a disorder of cell growth in the skin. has big impact, causing physical disabilities that can be seen directly and high treatment costs. In addition, skin also causes death if nor treated properly. Generally, dermatologists diagnose presence human body by using Biopsy process. this study, Dilated Convolutional Neural Network method was used to classify image data. development modifying dilation factors. The divided into two stages, including feature extraction fully connected layer. data study HAM1000 dataset. are dermoscopic datasets which consists 10015 images from 7 types cancer. This conducted several experimental scenarios changes value d, 2,4,6, 8 get optimal results. parameters epoch = 100, minibatch size 8, learning rate 0.1, dropout 0.5. best results were obtained with d=2 accuracy 85.67% sensitivity 65.48%.