SLSDeep: Skin Lesion Segmentation Based on Dilated Residual and Pyramid Pooling Networks

Skin lesion segmentation (SLS) in dermoscopic images is a crucial task for diagnosis of melanoma. In this paper, we present a robust deep SLS model, so-called SLSDeep, which is represented as an-decoder network. The encoder network is constructed by dilated residual, in turn, a pyramid pooling network followed by three convolution layers used for the decoder. Unlike the traditional methods employing a-entropy loss, we investigated a loss function by combining both Negative Likelihood (NLL) and End Point Error (EPE) to accurately segment the regions with sharp boundaries. The robustness of the proposed model evaluated on two public databases: ISBI 2016 and 2017 for skin lesion towards melanoma detection challenge. The proposed model outperforms state-of-the-art methods in terms of segmentation accuracy. Moreover, it is to segment more than $100$ images of size 384x384 per second on a GPU.