Dataset Overview:

• 80 of the images in the RGB retina dataset (https://www.kaggle.com/datasets/animalbiometry/cattle-retinal-fundus-images) obtained from the previous study were manually segmented by experts.
• A Samsung tablet and tablet pen were used during the segmentation process.
• The size of the segmented images is 1536×1152 pixels.
• Manually segmented retinal images were used to train the deep learning models and the performance of the models was tested with 1200 RGB images.
• The images used for the test were taken from the RGB images available at: (https://www.kaggle.com/datasets/animalbiometry/cattle-retinal-fundus-images).

Example manually segmented retinal images used in model training:

Sample RGB retinal images used for testing:

Citation
If you use the Cattle Retinal Fundus Images, please cite the dataset paper:
Ahmet Saygılı, Pınar Cihan, Celal Şahin Ermutlu, Uğur Aydın, Özgür Aksoy,
CattNIS: Novel identification system of cattle with retinal images based on feature matching method,
Computers and Electronics in Agriculture, 221 (2024): 108963, https://doi.org/10.1016/j.compag.2024.108963.

Funding:

• This project was supported by the Turkish Scientific and Technical Research Council-TÜBİTAK (Project Number: 121E349).

Ethics Committee Approval:

• The study was approved from the Kafkas University Animal Experiments Local Ethics Committee (Protocol number: KAU-HADYEK/2021-132).
• The study was approved from the Kafkas University Animal Experiments Local Ethics Committee (Protocol number: KAÜ-HADYEK/2024-123).