Airport Foreign Object Small Target Detection Dataset

To address the current issues in the field of airport runway foreign object detection, such as small data volume, low data quality, and limited data variety, a road foreign object dataset is proposed in this paper. The dataset is created through large-scale data collection and annotation, aiming to simulate the foreign object detection scenarios on airport runways and construct a diverse and detailed road foreign object dataset. This dataset consists of 12 categories of foreign objects, including metal devices, mechanical tools, and miscellaneous items. Furthermore, the dataset proposed in this paper was trained and tested using commonly used object detection algorithms. The experimental results demonstrated the significance of preprocessing the original dataset, and the Mean Average Precision(mAP) metric of the YOLO-V5-n algorithm cited in this paper outperformed other algorithms. The road debris dataset proposed in this paper provides a valuable database for the airport runway debris detection field.