Embedded systems are used in precision agriculture for data collection via sensors and for the control of actuators such as sprayers based on machine learning models. For plant classification and monitoring, it is easier to collect data of healthy plants than it is to collect data of plants that are infected by various diseases, because they are simply more common. Sufficient data are therefore often lacking for the accurate detection of diseased plants. In this paper, we outline an approach for the generation of synthetic data of infected plants that can be used to train a machine learning model for the classification of sugar beets. We use image augmentation techniques to build a pipeline that can automatically overlay diseased areas on healthy areas of leaf images.