CFP Dataset

该数据集包含了25个子集，共收录了103,786张彩色眼底摄影（CFP）图像，并由专家进行了标注，涵盖了广泛的眼科疾病。此外，这些数据被用于训练UrFound模型，以学习视网膜图像分析的表达式。该数据集的规模达到了103,786张图像，任务旨在进行表示学习以及诊断分类任务的预训练。This dataset comprises 25 subsets, containing a total of 103,786 color fundus photography (CFP) images. All images were expertly annotated and cover a wide range of ophthalmic diseases. Furthermore, this dataset was used to train the UrFound model to learn expressive representations for retinal image analysis. Boasting a total of 103,786 images, this dataset is intended for representation learning and pre-training for diagnostic classification tasks.

搜集汇总

背景与挑战

背景概述

CFP Dataset是UrFound项目的一部分，专注于通过多模态视网膜图像（如CFP和OCT）和专家注释的文本监督学习通用视网膜基础模型。该数据集支持知识引导的掩码建模技术，旨在提高模型在视网膜图像分析任务中的泛化能力和数据效率。