A maize leaf image dataset with three water treatment levels
A novel maize leaf data set is constructed, which contains 18040 images under three water treatments (well-watered, reduced-watered and drought-stressed). These local leaf images are cropped from 656 original maize digital images under natural environment. On the one hand, maize water stress status can be detected through exploring leaf phenotypic traits, like color changes, curling and wilting. On the other hand, a mass of image data is augmented that helps a lot to train a CNN.