UAVDT (Unmanned Aerial Vehicle Benchmark Object Detection and Tracking)

“UAVDT 是一个大规模具有挑战性的无人机检测和跟踪基准（即，来自 10 小时原始视频的大约 80, 000 个代表性帧），用于 3 个重要的基础任务，即对象检测 (DET)、单对象跟踪 (SOT)和多目标跟踪（MOT）。数据集由无人机在各种复杂场景中捕获。该基准中感兴趣的对象是车辆。这些帧是用边界框和一些有用的属性手动注释的，例如车辆类别和遮挡。 UAVDT 基准由 100 个视频序列组成，这些视频序列是从使用无人机平台在市区多个地点拍摄的超过 10 小时的视频中选出的，代表各种常见场景，包括广场、主干道、收费站、高速公路、十字路口和 T-视频以每秒 30 帧 (fps) 的速度录制，JPEG 图像分辨率为 1080 × 540 像素。"

UAVDT is a large-scale, challenging unmanned aerial vehicle (UAV) detection and tracking benchmark that contains approximately 80,000 representative frames extracted from over 10 hours of raw video, targeting three core fundamental tasks: object detection (DET), single-object tracking (SOT), and multi-object tracking (MOT). The dataset was captured by UAVs in various complex scenarios. The objects of interest in this benchmark are vehicles. These frames are manually annotated with bounding boxes and several useful attributes such as vehicle category and occlusion status. The UAVDT benchmark consists of 100 video sequences selected from over 10 hours of footage shot at multiple urban locations using UAV platforms, which cover various common scenarios including squares, main roads, toll stations, highways, intersections, and T-junctions. The videos were recorded at 30 frames per second (fps), with JPEG images having a resolution of 1080 × 540 pixels.

搜集汇总

背景与挑战

背景概述

UAVDT是一个大规模无人机检测和跟踪基准数据集，包含约80,000帧来自10小时无人机视频的数据，覆盖复杂场景如广场和高速公路，专注于车辆对象的检测和跟踪任务。数据集提供手动注释的边界框、车辆类别和遮挡属性，支持对象检测、单对象跟踪和多目标跟踪研究，适用于计算机视觉领域。