UAVDT
收藏资源简介:
UAVDT数据集由中国科学院大学等机构创建,包含约80,000帧从10小时无人机拍摄视频中精选的图像,覆盖多种复杂城市环境。数据集主要关注车辆目标,每帧均标注了边界框及多达14种属性,如天气条件、飞行高度、相机视角等。该数据集旨在推动无人机视觉技术在不受限制场景下的研究,解决高密度、小目标、相机运动等挑战,适用于物体检测、单目标跟踪和多目标跟踪等基础视觉任务。
UAVDT dataset was developed by the University of Chinese Academy of Sciences and other institutions. It contains approximately 80,000 frames of carefully selected images extracted from 10 hours of drone-captured videos, spanning a wide range of complex urban environments. The dataset primarily focuses on vehicle targets, with each frame annotated with bounding boxes and up to 14 attributes including weather conditions, flight altitude, camera perspective, and others. This dataset aims to advance the research of unmanned aerial vehicle vision technology in unconstrained scenarios, addressing challenges such as high-density scenes, small-sized targets, and camera motion, and is suitable for fundamental computer vision tasks including object detection, single-object tracking, and multi-object tracking.

- UAVDT数据集首次发表,由武汉大学和香港理工大学联合发布,旨在为无人机视频中的目标检测和跟踪提供一个标准化的基准。
- UAVDT数据集首次应用于国际计算机视觉与模式识别会议(CVPR)的无人机目标检测挑战赛,推动了无人机视频分析技术的发展。
- UAVDT数据集被广泛应用于多个研究项目和学术论文中,成为无人机视频分析领域的重要参考数据集。
- UAVDT数据集的扩展版本发布,增加了更多的视频样本和目标类别,进一步丰富了数据集的内容和多样性。



