Eye Conjunctiva Segmentation Dataset

The dataset consists of 547 eye images showing conjunctiva collected using two smartphones, the OnePlus 9R(48 MP, f/1.7) and OnePlus 9Pro(48 MP, f/1.8). The room where the images were captured had an adequate amount of lighting. While taking images the lower eyelid was gently moved down with the thumb to get accurate exposure to the conjunctiva of the eyes. No additional lighting was used during the image acquisition process.

All the images were labeled using “Labelme” software by two annotators. Inter Annotator Agreement (IAA) was calculated among all the labeled images as an intersection percentage between annotated images from two annotators. The agreement found was almost perfect which is on average about 99.9%. The images are of very high resolution which is around (3000X4000). Annotated masks are of the same resolution as the corresponding image. The dataset contains three files. The file named 'Images' contains 547 original images showing eye conjunctiva. Files named 'Masks Annotator 1' and 'Masks Annotator 2' contain masks annotated by two different annotators.