无人机视觉识别数据集
收藏资源简介:
AU-AIR是首个用于目标检测的多模态无人机视觉识别数据集,包含超过2小时原始视频、32,823张标注帧和132,034个目标实例,涵盖8类交通监控相关目标。数据集融合视觉图像、精确对象标注及飞行数据(如时间、GPS、IMU、高度和速度),支持空中图像中的目标检测与定位。其多模式特性满足自主空中监控需求,适用于无人机测绘、建筑检查等场景,推动计算机视觉与机器人算法在复杂空域环境中的研究与应用。
AU-AIR is the first multi-modal drone visual recognition dataset for object detection. It contains over 2 hours of raw video, 32,823 annotated frames, and 132,034 object instances, covering 8 categories of traffic monitoring-related objects. This dataset integrates visual images, precise object annotations, and flight data including time, GPS, IMU, altitude and speed, enabling object detection and localization in aerial images. Its multi-modal characteristics meet the requirements of autonomous aerial surveillance, and is applicable to scenarios such as UAV mapping and building inspection, promoting the research and application of computer vision and robotic algorithms in complex airspace environments.




