Dataset of GrowScreen-Rhizo 3 shoot image analysis including trained U-Net CNN

10.26165/JUELICH-DATA/SOJ6SQ Constantin Eiteneuer0000-0003-3170-3561(Forschungszentrum Juelich - IBG-2)Laura Verena Junker-Frohn0000-0002-0655-6232(Forschungszentrum Juelich - IBG-2)Henning Lenz0000-0002-8080-0328(Forschungszentrum Juelich - IBG-2)Mark Müller-Linow0000-0002-6722-5437(Forschungszentrum Juelich - IBG-2)Kerstin A. Nagel0000-0003-3025-0388(Forschungszentrum Juelich - IBG-2) Dataset of GrowScreen-Rhizo 3 shoot image analysis including trained U-Net CNN Jülich DATA 2026 The publication contains 1800 RGB images generated using the high throughput plant phenotyping platform GrowScreen-Rhizo 3. Shoots of barley, maize, and sunflower plants differing in plant developmental stages, with a maximum age of 3-4 weeks, were imaged by 6 cameras from 3 different view points (top view, 45°, side view) during multiple rhizotron experiments (for details, see Junker-Frohn et al., 2026). 600 images each for barley, maize, and sunflower plants and their corresponding manually annotated masks were split 5:1 for model evaluation and validation purposes. The pytorch U-Net implementation that was trained on the data for inference is included in this publication. Furthermore the publication contains 696 RGB images, imaged with the GrowScreen-Rhizo3 platform, and their corresponding measured leaf area and plant height. This data was used to train a Gaussian Process Regression model to predict leaf area. The trained model for inference is included in this publication. Constantin Eiteneuer(IBG-2: Plant Sciences, Forschungszentrum Jülich GmbH)