10.7910/DVN/N3YJXGWang, RichardRichardWangHarvard Dataverse2023Agricultural SciencesComputer and Information ScienceWang, RichardRichardWang2023-03-122023-03-132147483648214748364821474836481505278551144770application/octet-streamapplication/octet-streamapplication/octet-streamapplication/ziptext/x-shapplication/octet-stream1.0CC0 1.0A dataset for the semantic segmentation of aphid clusters in sorghum fields. The dataset is made up of 54,742 images and their corresponding masks. From initial high-resolution images, we randomly shuffled and split the images into 10 separate groups from which we generated patches at three separate scales with a 10% overlap. The three scales 0.132Hx0.132W, 0.263Hx0.263, and 0.525Hx0.525W have 36478, 14628, and 3636 images respectively.