Cassava root cross-section images

Published: 31-07-2020| Version 3 | DOI: 10.17632/gvp7vshvnh.3
Joyce Nakatumba-Nabende,
Ernest Mwebaze,
Robert Kawuki,
Heneriko Kulembeka,
Solomon Nsumba,
Jeremy Tusubira,
Samiiha Nalwooga,
Benjamin Akera,
Ruhinda Kabonire,
Daniel Ssendiwala,
Paul Mugisha


This dataset contains images of cassava root cross-sections captured by the Makerere University Artificial Intelligence Lab in conjunction with the National Crop Resources Research Institute in Uganda and the Tanzania Agricultural Research Institute in Tanzania. The images were captured during harvests conducted by cassava breeders to assess and score root necrosis for different varieties of cassava. The dataset contains 10052 images from five field trials. The images can be used to develop deep learning algorithms that can help to automate the scoring of root necrosis and to study other aspects of necrosis expression on cassava roots. We have also added a ground truth dataset of 1036 images and corresponding masks that can be used for image analysis experiments.