2026-10-14T08:21:44Z https://www.scidb.cn/oai

10.57760/sciencedb.j00240.00278 2026-08-05T17:24:06Z

2026-08-05 A Multi-source Dermoscopic Image Dataset for Skin Lesion Segmentation doi:10.57760/sciencedb.j00240.00278 en This dataset is designed for automated skin lesion segmentation and is compiled from three public dermoscopic image datasets: ISIC2017, ISIC2018, and PH2. It contains 5,044 dermoscopic images and their corresponding lesion masks, including 2,150 images from ISIC2017, 2,694 images from ISIC2018, and 200 clinically annotated images from PH2. The dataset covers multiple skin lesion categories, including melanoma, benign nevus, atypical nevus, and seborrheic keratosis. It exhibits substantial variations in lesion scale, color, texture, and boundary clarity, while also containing challenging artifacts such as hairs and specular reflections. For deep learning-based training and evaluation, the images and masks can be normalized and resized to a unified resolution, and randomly divided into training and test sets at a ratio of 7:3. The dataset supports research on skin lesion segmentation, medical image semantic segmentation, lightweight model evaluation, and boundary-aware segmentation methods. skin lesion segmentation; dermoscopic images; medical image segmentation; semantic segmentation; lesion masks; multi-source dataset liu yang Qiu Yunfei Han Xuefeng PUBLIC https://creativecommons.org/licenses/by-nc-nd/4.0/ dataset Science Data Bank