Skin Cancer Classification for Dermoscopy Images Using Convolutional Neural Networks.

Abstract

Deep Learning model gives remarkable result for automated skin lesion analysis. To Train Deep Learning Model it requires considerable high amount of data. But annotated image is available in very limited amount By Technique of data augmentation training data can be extended through transforming images. In this work I have closely went through the all augmentation scenarios for skin data and have done classification using CNN architecture to classify the melanoma and non melanoma without any augmentation, that gives the AUC value 0.861 and precision is 94, recall 77, F1 score 85.