Rotterdam artery-vein segmentation (RAV) dataset
A new dataset of 1024x1024-pixel retinal images includes challenging, real-world samples often excluded by automated quality checks.
Researchers from the Rotterdam Study and Eyened Reading Center released the Rotterdam Artery-Vein (RAV) Segmentation dataset. It contains high-resolution (1024x1024px) color fundus images with manually verified artery/vein masks, including difficult real-world samples. The dataset provides three image modalities: original, contrast-enhanced, and segmentation masks. This enables developers to train and benchmark more robust computer vision models for automated retinal analysis, improving tools for disease screening and diagnosis in ophthalmology.
Why It Matters
Better training data leads to more reliable AI for detecting systemic diseases like diabetes and hypertension from retinal scans.