The SUSTech-SYSU dataset for automatically segmenting and classifying corneal ulcers

In this zip folder, all the original ocular staining images, the associated cornea labels, and the associated flaky ulcer segmentation labels (images with point-like corneal ulcers do not have such segmentation labels) are provided as three folders, named respectively as "rawImages", "corneaLabels", and "ulcerLabels" (Data Citation 1). Results of superimposing cornea and flaky corneal ulcer segmentation masks on the associated ocular staining images are also provided in the "corneaOverlay" and "ulcerOverlay" folders. In the "rawImages" folder, files are named as "n.jpg", with n ranging between 1 and 712 denoting the nth sample. All images in this folder are saved in the JPG format. The segmentation label images of the cornea and the flaky ulcer area are saved as binary images in the PNG format, having the same size as the corresponding ocular staining image. An excel file was also provided, with the first column denoting the sample ID, the second column denoting the general pattern category (0 to 2, with 0 representing a point-like corneal ulcer, 1 representing a point-flaky mixed corneal ulcer, and 2 representing a flaky corneal ulcer), the third column denoting the type grading (0 to 4), and the fourth column denoting the grade grading (0 to 4). Two videos respectively demonstrating the semi-automatic segmentation process and the subsequent manual correction process were also provided in the zip folder.