DeepSkinNet: A Deep Learning Model for Skin Cancer Detection
Alla Abhiram, S. M. Anzar, Alavikunhu Panthakkan
Abstract
In this article, an intelligent skin cancer detection system is proposed. It is common knowledge that skin cancer is one of the deadliest diseases in the world. Therefore, in order to save human lives, it must be detected correctly at the early stage, for which a fully automatic system using Deep Learning techniques can be used. In this work, the HAM10000 dataset is used for skin cancer image classification. A specific model for skin cancer classification, DeepSkinNet model, is proposed and tested. For skin lesion classification, the proposed DeepSkinNet model is compared to contemporary models such as AlexNet, VGG-16, and InceptionV3. The same dataset is trained with all the above models and the confusion matrix of the models is determined. Known quantitative measures such as accuracy, precision and recall are also evaluated. Finally, the experimental studies shows that the proposed model achieves better accuracy than the other models. The DeepSkinNet model has an accuracy, precision, recall and F1-score of 97.354%, 98%, 97% and 97% respectively.