AirBirds
收藏资源简介:
AirBirds是一个专为机场鸟击预防设计的大型挑战性数据集,由中国人民大学创建。该数据集包含118,312张时间序列图像,总计409,967个飞行鸟类的边界框标注,图像分辨率为1920x1080,平均标注实例大小小于10像素。数据集通过在真实机场部署的摄像头网络,覆盖了一整年的四季,记录了多种鸟类、光照条件和13种气象场景。创建过程历时两年,涉及数据收集、预处理、标注和分割等多个阶段。AirBirds旨在解决机场鸟击问题,通过提供真实世界机场中的飞行鸟类数据,推动鸟击预防技术的研究与应用。
AirBirds is a large-scale challenging dataset specifically designed for airport bird strike prevention, created by Renmin University of China. This dataset includes 118,312 time-series images, with a total of 409,967 bounding box annotations for flying birds. The image resolution is 1920×1080, and the average size of each annotated instance is less than 10 pixels. The dataset was gathered through a network of cameras deployed at real-world airports, covering all four seasons across a full year, and records diverse bird species, varying lighting conditions, and 13 types of meteorological scenarios. The entire creation process spanned two years, involving multiple stages such as data collection, preprocessing, annotation, and dataset splitting. AirBirds aims to tackle the airport bird strike issue, and advance the research and application of bird strike prevention technologies by providing real-world flying bird data collected from actual airport environments.

- 1AirBirds: A Large-scale Challenging Dataset for Bird Strike Prevention in Real-world Airports中国人民大学 · 2023年



