Representative dataset for machine learning-based RGB image analysis of barley foliar disease resistance (detached leaf assay phenotyping)

This dataset provides a representative subset of the detached leaf assay (DLA) imagery obtained using Macrobot Nextgen and derived quantitative trait data used to develop and validate the RGB image analysis pipeline described in the manuscript. Pixel classification was performed using a shallow feedforward artificial neural network (10 hidden neurons, single hidden layer, 4-class output; MATLAB Neural Network Pattern Recognition App, R2022b), trained to classify image pixels into four classes — background, necrotic tissue, chlorotic tissue, and healthy green tissue — achieving 94.1% pixel classification accuracy on held-out validation data. A total of 113 features were derived from image analysis per leaf sample from the classified pixel maps, including main quantitative trait values (symptom-based percentages, RGB band averages, and derived vegetation indices). Please refer to README file for details.