2026-10-15T02:02:07Z https://www.scidb.cn/oai

10.57760/sciencedb.37353 2026-05-26T09:15:58Z

2026-05-26 GrainPest9 doi:10.57760/sciencedb.37353 en We constructed a dedicated granary pest dataset, GrainPest9, containing 2,278 images and 7,474 annotated instances from nine representative stored-grain pest species. Collected from real grain-storage scenarios, GrainPest9 better reflects practical challenges such as complex backgrounds, small object sizes, and inter-class similarity, providing a solid data basis for intelligent granary pest monitoring.## Data File DescriptionThe GrainPest9 dataset is organized in YOLO annotation format. Each image has a corresponding annotation file with the same base name. For example, the image file 000001.jpg corresponds to the annotation file 000001.txt.The dataset mainly contains two types of files: image files and annotation files. The image files store the original pest images and can be used for training, validation, and testing of object detection models. The annotation files record the category information and bounding-box location of pest targets in each image.Each image has a corresponding .txt annotation file. Each row in the annotation file represents one pest instance. If an image contains multiple pest targets, the corresponding annotation file contains multiple rows. Each row follows the YOLO annotation format:class_id x_center y_center width heightHere, class_id represents the pest category ID; x_center represents the normalized x-coordinate of the bounding-box center; y_center represents the normalized y-coordinate of the bounding-box center; width represents the normalized width of the bounding box; and height represents the normalized height of the bounding box. All coordinate values are normalized to the range of 0 to 1 according to the image width and height. Therefore, the bounding-box coordinates are independent of the original image resolution and can be directly used for training YOLO-style object detection models.For example, one annotation row is shown as follows:0 0.5123 0.4387 0.0865 0.0742This row indicates that the pest instance belongs to category 0, and its bounding box is represented by the normalized center coordinates, width, and height.The GrainPest9 dataset contains nine pest categories. The category IDs are consistent with the names field in the dataset configuration file data.yaml. The category mapping is as follows:0: Sitophilus oryzae1: Oryzaephilus surinamensis2: Rhyzopertha dominica3: Tribolium castaneum4: Cryptolestes ferrugineus5: Aphididae6: Sitobion avenae7: Bothynoderes punctiventris8: Haplothrips triticiGrainPest9 can be directly used for training, validation, and testing of YOLO-format object detection models. This dataset is especially suitable for evaluating small-object pest detection performance in granary-related scenarios, such as complex backgrounds, grain kernel interference, impurity occlusion, scale variation, and trap-like imaging environments.Compared with general agricultural pest datasets, GrainPest9 focuses more on stored-grain pest monitoring. It provides image and annotation data that are closer to real granary environments, offering data support for the development of lightweight and real-time pest detection models in intelligent grain storage management. Stored-grain pests; Small-object detection; Pest image annotation Yan EMBARGO https://creativecommons.org/licenses/by/4.0/ dataset Science Data Bank