Fundus Image Segmentation Benchmark Dataset for Blood Vessel Segmentation

This benchmark dataset was created by combining three widely-used datasets: CHASEDB1, STARE, and DRIVE, to evaluate and enhance the performance of blood vessel segmentation models in fundus images. The dataset includes 28 high-resolution images from CHASEDB1, 20 diverse images from STARE, and 40 images from DRIVE, all preprocessed for uniformity in terms of size, contrast, and illumination. The data was split into training (70%), validation (15%), and testing (15%) sets, ensuring balanced representation for model development and unbiased evaluation. This comprehensive benchmark dataset has been openly published to aid in the development and comparison of models for robust fundus vessel segmentation.

本基准数据集由CHASEDB1、STARE与DRIVE三个广泛使用的数据集融合构建，旨在评估并优化眼底图像血管分割模型的性能。该数据集包含CHASEDB1的28张高分辨率图像、STARE的20张多样化图像以及DRIVE的40张图像，所有图像均已针对图像尺寸、对比度与光照条件完成统一预处理。数据集按70%训练集、15%验证集、15%测试集的比例划分，确保模型开发与无偏评估所需的数据分布均衡。本综合基准数据集已开源发布，旨在助力鲁棒性眼底血管分割模型的研发与性能对比工作。