Melanoma Detection Using Deep Learning

In this paper, we describe a region of interest-based approach for the classification of dermoscopic images of skin lesions, which nowadays contributes to early identification of skin melanoma. Once the region of interest detected, it will be further processed in order to be used for training and hence classification using deep learning methods. The main goal is to compare three different convolutional neural networks (CNNs) models and determine the one which provides the best accuracy, knowing that only salient parts of the skin lesions images have been used for training.