PBDet-Soccer: An annotated dataset for soccer player and ball detection

PBDet-Soccer is an annotated image dataset for soccer player and ball detection in match images. The dataset contains 11,800 images and two object categories: player and ball. Bounding-box annotations are provided in YOLO normalized format. The archive contains the training, validation, and test image subsets; the corresponding YOLO annotation folders; and a data.yaml configuration file. Class identifiers and dataset paths are defined in data.yaml. The dataset was prepared for evaluating lightweight object detection methods for soccer match analysis, including player detection, small-ball detection, and model efficiency assessment.