Evaluation benchmark for natural robustness evaluation of retinal vessel segmentation models

<p>A dataset contains benchmark images for natural robustness evaluation of deep learning models for retinal vessel segmentation. The dataset consists of three mainstream retinal vessel segmentation datasets: DRIVE, STARE, and CHASE_DB1.</p> <p>For each dataset are provided:</p> <ul> <li><em>images </em>- directory containing fundus images augmented using <a href="https://github.com/goranagojic/AugOOD">AugOOD</a> tool for fast image augmentation for OOD robustness evaluation.</li> <li><em>labels</em> - directory with labels that correspond to the images.</li> <li><em>masks</em> - directory with FoV masks that correspond to the images.</li> </ul> <p>The benchmark is used in the paper <a href="https://onlinelibrary.wiley.com/doi/abs/10.1002/cpe.6809">Robustness of deep learning methods for ocular fundus segmentation: Evaluation of blur sensitivity</a> to evaluate natural robustness of a portfolio of deep learning models for retinal vessel segmentation from fundus images.</p>