Ensembles of Deep Convolutional Neural Networks for Detecting Melanoma in Dermoscopy Images

Malignant melanoma is the deadliest form of skin cancer and is one of the most rapidly increasing cancers in the world. In this paper, a methodology for the SIIM-ISIC Melanoma Classification Challenge, where the goal is to detect melanoma from dermoscopic images, is described. The EfficientNet family of convolutional neural networks is utilized and extended for identifying malignant melanoma on a dataset of 58,457 dermoscopic images of pigmented skin lesions. This binary classification problem comes with a severe class imbalance, which is tackled using a loss balancing approach. Furthermore, the dataset contains images with different resolution sizes. This property is addressed by considering different model input resolutions. Lastly, an ensembling strategy of models, trained with different activation functions is applied to increase the diversity of the ensembler and to further improve individual results.