iChallenge-GON

iChallenge-GON是1200个带注释的视网膜眼底图像的大型数据集。 根据1:1:1比例将数据集分为3个子集，分别用于训练、离线验证 (offline validation) 和现场测试 (onsite test)，每个子集中青光眼的比例相同。 训练集由400彩色眼底图像组成，其将提供相应的青光眼状态和统一的手动像素标注 (也就是地面真相)。测试集由800彩色眼底图像组成，其中400是非现场验证集图像，400是现场测试集图像。iChallenge-GON is a large-scale dataset containing 1200 annotated retinal fundus images. The dataset is split into three subsets at a 1:1:1 ratio for training, offline validation, and onsite test respectively, with each subset holding an identical proportion of glaucoma cases. The training set comprises 400 color fundus images, accompanied by corresponding glaucoma status labels and uniform manual pixel-level annotations (i.e., ground truth). The test set consists of 800 color fundus images, among which 400 are offline validation set images and 400 are onsite test set images.

搜集汇总

背景与挑战

背景概述

iChallenge-GON是一个包含1200张带注释视网膜眼底图像的数据集，按1:1:1比例分为训练、离线验证和现场测试三个子集，每个子集400张图像且青光眼比例相同。训练集提供青光眼状态和手动像素标注，测试集包括非现场验证和现场测试图像。该数据集由百度于2018年发布。