RAVIR: A Dataset and Methodology for the Semantic Segmentation and Quantitative Analysis of Retinal Arteries and Veins in Infrared Reflectance Imaging




The retinal vasculature provides important clues in the diagnosis and monitoring of systemic diseases including hypertension and diabetes. The microvascular system is of primary involvement in such conditions, and the retina is the only anatomical site where the microvasculature can be directly observed. The objective assessment of retinal vessels has long been considered a surrogate biomarker for systemic vascular diseases, and with recent advancements in retinal imaging and computer vision technologies, this topic has become the subject of renewed attention. In this paper, we present a novel dataset, dubbed RAVIR, for the semantic segmentation of Retinal Arteries and Veins in Infrared Reflectance (IR) imaging. It enables the creation of deep learning-based models that distinguish extracted vessel type without extensive post-processing.

For more information, please refer to the dataset paper:

RAVIR: A Dataset and Methodology for the Semantic Segmentation and Quantitative Analysis of Retinal Arteries and Veins in Infrared Reflectance Imaging


Data Download

Please see the Data page for downloading the RAVIR dataset. 


Test Set Evaluation

Note that the following needs to be taken into account when preparing the prediction files:

  • Predictions should be a PNG file as 2D maps containing artery and vein classes with size (768,768) [ submitting maps with size (768,768,3) will trigger an error in evaluation).
  • Artery and vein classes should have labels of 128 and 256 respectively. Background should have a label of 0
  • The filenames must exactly match the name of the images in the test set [IR_Case_006.png, ..., IR_Case_060.png].

Once segmentation outputs are obtained in the correct format, participants should place the them in a folder named test and submit a zipped version of this folder to the server. 

The Dice and Jaccard scores will be calculated for every image in the test set for both artery and vein classes. The leaderbord is sorted on the basis of the best average Dice score. 


Note: If you previously downloaded the dataset ( before Aug 16, 2022), please remove case IR_Case_002.png from test set images and predictions as it is a duplicate image. Failure to do so will result in errors in getting the segmentation performance from the test set server. There should be 19 cases in test set in total. Nothing needs to be changed if the dataset is downloaded after this date.