AgriUnion: A Unified Multi-Source UAV Agricultural Dataset for Tiny Object Detection

AgriUnion is a multi-source unified agricultural aerial dataset designed for tiny object detection and UAV edge deployment. It contains 8,864 high-resolution UAV images with 317,704 annotated bounding boxes, covering 29 categories of crops and weeds in complex field scenarios. The annotation follows standard YOLO format, and the dataset is split into training/validation/test sets with a 7:2:1 ratio via stratified sampling. 74.4% of instances are COCO-defined small objects (<32×32 pixels). This dataset provides a unified benchmark for agricultural intelligent perception and lightweight detection model research.