Digitised Herbarium Image Segmentation Dataset

This dataset was created to support the training and evaluation of semantic segmentation models for plant region extraction in digitised herbarium specimens. It was created in the context of the e-col+ project (ANR‐21‐ESRE‐005).<b>Content</b><b>Segmentation dataset:</b>The dataset consists of original digitised herbarium specimen images and their corresponding segmented versions, in which only the plant regions are preserved while background elements (labels, barcodes, colour charts) are masked. Images are split into two main categories:<code>unsegmented_images/</code>: original RGB images<code>segmented_images/</code>: processed images with background removed manually. Each of these directories contains three subfolders corresponding to the standard data splits:<code>train/</code>: 1,180 images<code>val/</code>: 296 images<code>test/</code>: 333 imagesTraining and validation images were selected from the <b>Herbarium Segmentation Dataset </b>(https://doi.org/10.6084/m9.figshare.27685914.v1), while the test set consists of newly and manually annotated images.<b>Out-of-Distribution Evaluation dataset:</b>The Out Of Distribution (OOD) dataset comprises 171 unannotated images, selected to represent visually challenging conditions, including:noisy or highly textured backgrounds,colored backdrops (e.g., yellow, pink, dark grey),intricate plant morphologies,and common digitisation artifacts like pins, overlapping components, and mosaic-like patterns.