Amir Hossein Saleknia - Master student, Iran University of Science and Technology
Ahmad Ayatollahi - Professor, Iran University of Science and Technology

During the last decade, convolutional neural networks have made more precise predictions for medicalimage segmentation problems. Nevertheless, these networks mostly require high computational capabilityand massive storage, which prevent their application in real-world situations. To address these issues, wepropose an efficient skin segmentation network, named ESS-Net, for automated skin lesion segmentationfrom dermoscopic images. Results demonstrate that the proposed method outperforms state-of-the-art deeplearning models on the ISIC ۲۰۱۷ dataset despite of its limited computational requirements.

Deep learning, Convolutional neural networks, Skin lesion segmentation