STARE STructured Analysis of the Retina Dataset

This package is part of the signal-processing and machine learning toolbox Bob. It provides an interface for the STARE Dataset. This package does not contain the original data files, which need to be obtained through the link above.

The dataset contains 20 eye fundus images with a resolution of 700 x 605. Two sets of ground-truth vessel annotations are available. The first set by Adam Hoover is commonly used for training and testing. The second set by Valentina Kouznetsova acts as a “human” baseline.

If you use this package, please cite the authors of the database:

@ARTICLE{845178,
    author={A. D. {Hoover} and V. {Kouznetsova} and M. {Goldbaum}},
    journal={IEEE Transactions on Medical Imaging},
    title={Locating blood vessels in retinal images by piecewise threshold probing of a matched filter response},
    year={2000},
    volume={19},
    number={3},
    pages={203-210},
    doi={10.1109/42.845178},
    ISSN={0278-0062},
    month={March},
}