Dataset for Root Segmentation Beyond Species Boundaries: A Generalizable Framework for Anatomical Analysis

Code available at https://github.com/janetkok/Root-Segmentation-Beyond-Species-BoundariesDataset Overview This dataset provides annotated imagery for the semantic segmentation of Pearl Millet and Sorghum crops. It is designed to support both binary and multiclass segmentation machine learning tasks. Repository Structure The dataset is organized into two main distinct directories: binary/: Designed for binary segmentation models. Content: Pearl Millet images. Annotations: Ground truth binary segmentation masks (Foreground/Background). multiclass/: Designed for multiclass segmentation models. Content: Images of both Pearl Millet and Sorghum. Annotations: Ground truth multiclass segmentation masks. Experimental Splits To ensure consistent evaluation, standard Training and Validation splits are provided for: Pearl Millet only. Sorghum only. Combined Dataset (Pearl Millet + Sorghum). FundingThis work was supported by the Biotechnology and Biological Sciences Research Council (grant number BB/Y513908/1) and Agence National de la Recherche (SorDrought grant ANR-23-CE20-0052).