Full-Reference Evaluation Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/Retinal-Research/EyeBench
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含16,817张高质量的眼底图像，这些图像被分为训练、验证和测试三个部分，专门用于视网膜图像增强任务。此外，这些图像按照三个质量等级进行分类，并标注了糖尿病视网膜病变的严重程度标签。数据规模为16,817张图像，其中10,000张用于训练，600张用于验证，剩余的6,217张用于测试。该数据集的任务是进行视网膜图像的增强和评估。
This dataset contains 16,817 high-quality fundus images, which are split into training, validation and test subsets specifically designed for retinal image enhancement tasks. Furthermore, these images are categorized into three quality grades and annotated with severity labels of diabetic retinopathy. Specifically, 10,000 images are allocated for training, 600 for validation, and the remaining 6,217 for testing. The core tasks of this dataset are retinal image enhancement and evaluation.
提供机构：
EyePACS initiative
