Plant Region Detection in Digitised Herbarium Specimens

This dataset was created as part of the e-col+ project (ANR‐21‐ESRE‐005) to support detection of plant regions in digitised herbarium sheets. It is intended for use in training and evaluating computer vision models for plant detection and segmentation tasks.
Content
The dataset consists of digitised herbarium specimen images annotated with bounding boxes identifying the vegetative regions (i.e., plant material). It consists of 19,078 image patches annotated with bounding boxes. Images were collected from the ReColNat website (https://recolnat.fr) and split into:
14,307 training samples3,815 validation samples956 test samplesThe dataset includes two main directories: images/ and labels/, each subdivided into train/, val/, and test/folders. The images/ directory contains RGB .jpg files, while the labels/ directory includes annotation .txt files in YOLO format (class_id, x_center, y_center, width, height), with each label file named identically to its corresponding image.