imageomics/drone-maneuver-clips
收藏资源简介:
Drone Maneuver Test-Clip Library(无人机机动测试片段库)是一个包含41个六秒空中无人机视频片段的数据集,这些片段源自肯尼亚Mpala研究中心的KABR数据集,从20个KABR视频中切割而成,覆盖17次调查会话。每个片段根据其适合测试的自主飞行机动(如启动、跟随、行为自适应、兴趣状态感知)进行索引,并附带每帧每轨迹的标签表(包括边界框、物种、行为、持久轨迹ID、可用的真实姿态和遥测数据)。该库旨在评估无人机导航策略在真实野生动物视频上的性能,提供详细的标注信息,如物种(包括斑马、长颈鹿等)、行为、姿态和栖息地结构分类,适用于目标检测、跟踪、行为识别、姿态估计和机动条件导航策略评估等任务。
A library of 6-second aerial drone clips derived from the KABR dataset (Mpala Research Centre, Kenya), each indexed by the autonomous-flight maneuver it is suitable for testing, with per-frame-per-track labels (bounding box, species, behaviour, persistent track id, ground-truth pose where available, telemetry). A benchmark of 41 six-second drone clips (180 frames @ 30 fps) cut from 20 KABR videos across 17 survey sessions at the Mpala Research Centre, Kenya. Each clip is indexed by which autonomous-flight maneuver it can test (launch / follow / behavior-adaptive / SoI-aware) and ships with a per-frame-per-track label table. The library is intended for evaluating drone navigation policies against real wildlife footage.




