A dataset of field symptoms for identifying Citrus Huanglongbing

<p indent="0mm">The rapid identification of Citrus Huanglongbing can help producers to detect and eliminate infected plants as early as possible. The automatic recognition of plant diseases based on deep learning relies on large datasets, but there are few open datasets available for citrus huanglongbing recognition. In this study, a datasets of images was collected using a mobile phone or camera, involving three different collection environments: natural environment, white background, and black background. The dataset includes a total of 9,896 images including 14 categories of symptoms related to Huanglongbing and other associated diseases. Among these, 7,192 samples were used for the training set, 898 for the validation set, and 1,806 for the test set (of which 602 had PCR test results). The symptom categories of the samples were classified and annotated by plant detection experts to ensure the quality and reliability of the data. The datasets presents a YOLO format (txt) annotation file and model files trained using YOLOv5l, aiming to provide training and testing samples for developing models of machine learning based on the YOLO convolutional neural network in the fields of plant disease classification, object detection and automatic recognition.