Aberystwyth Leaf Evaluation Dataset

Abstract

We are releasing the dataset acquired to support the work of the EPSRC funded project “Dynamic Modelling of Plant Growth with Computer Vision” (grant code EP/LO17253/1) in the hope that its availability will help further advance the state of the art in the use of image analysis for plant sciences. It is a visible light, top down, timelapse image dataset of Arabidopsis thaliana (Arabidopsis). It incorporates the original images, some with leaf-level ground truth annotations, harvested plant ground truth data and scanned images together with supporting software. 4 sets of 20 Arabidopsis Thaliana plants have been grown in trays Images of each tray are taken in a 15 minute timelapse sequence using a robotic greenhouse system Periodically, plants from each tray are sacrificed to record destructive measurements A subset of these images have been hand-annotated to provide leaf-based image <em>ground truth</em>. In total, there are 56 annotated ground truth images containing 916 hand-marked up individual arabidopsis plants We also release software to help with the analysis of these images