Multimodal Object Detection Dataset for Car and Person Detection - Part 2

This multimodal dataset was acquired using a drone equipped with a DJI Zenmuse H20T camera to monitor a parking area by navigating through 5 predefined waypoints. The drone flew at an altitude of 60 meters and at each waypoint hovered for 30 seconds to record video using three sensors: the RGB wide lens, RGB zoom lens (at a minimum 2× zoom, simulating a 30-meter altitude), and the thermal lens. The recorded videos were processed to extract individual frames. The dataset is organized into separate ZIP archives for each waypoint and camera lens. Each archive is identified using a combination of the waypoint number and sensor type: 'W' for wide, 'Z' for zoom, and 'T' for thermal. Frame-level annotations are also included and organized in a similar manner.本多模态数据集由搭载DJI禅思H20T（DJI Zenmuse H20T）相机的无人机采集，该无人机通过沿5个预设航点飞行的方式对停车场开展监测。无人机飞行高度设定为60米，且在每个航点悬停30秒，通过三类传感器录制视频：RGB广角镜头、RGB变焦镜头（最低变焦倍数为2×，模拟30米高度下的拍摄效果）以及热成像镜头。所录制的视频经处理后提取出单帧图像。本数据集按每个航点与相机镜头类型分别打包为ZIP压缩档案，每个压缩档案以航点编号与传感器类型的组合命名：其中'W'代表广角镜头、'Z'代表变焦镜头、'T'代表热成像镜头。数据集同时包含帧级标注信息，且其组织形式与上述一致。