AeroIntent-landscape
收藏资源简介:
AeroIntent - Landscape Scene 是 AeroIntent 多模态同步基准数据集中的一个场景,专门用于无人机(UAV)意图识别任务。该数据集包含来自三个机身系列和五种有效载荷配置的15种无人机配置,以及22种轨迹程序,这些轨迹程序被归类为五种意图:Transit(过境)、Reconnaissance(侦察)、Probing(探测)、Strike(打击)和 Escape(逃脱)。所有数据仅从前方视角采集,提供了同步的多模态数据,包括 RGB 图像、红外图像、深度图像、Echo 雷达数据、IMU 数据、GPS 数据、边界框标注、时间戳和元数据。该场景共包含330个同步序列。数据通过 tar.zst 压缩分片存储,每个无人机配置对应一个分片,每个分片被分割成256 MiB的部分以便于可靠传输。用户可以通过提供的脚本重建并提取数据。数据集的目录结构清晰,每个无人机配置目录下按意图、轨迹组织,包含 rgb、ir、depth、depth_preview、echo、echo_groundtruth、imu、gps 等子目录,以及 labels_all.txt、timestamps.csv、meta.json 和多种视频文件。该数据集适用于视频分类、目标检测以及多模态意图识别等任务,总计样本数量在10万到100万之间。
AeroIntent - Landscape Scene is a scene within the AeroIntent multimodal synchronous benchmark dataset, specifically designed for Unmanned Aerial Vehicle (UAV) intention recognition tasks. The dataset includes 15 UAV configurations from three airframe series and five payload configurations, along with 22 trajectory programs classified into five intentions: Transit, Reconnaissance, Probing, Strike, and Escape. All data are collected from a forward-facing perspective, providing synchronized multimodal data including RGB images, infrared images, depth images, Echo radar data, IMU data, GPS data, bounding box annotations, timestamps, and metadata. This scene contains a total of 330 synchronous sequences. The data is stored in tar.zst compressed shards, each corresponding to one UAV configuration, with each shard split into 256 MiB parts for reliable transmission. Users can reconstruct and extract the data using provided scripts. The dataset directory structure is clear, with each UAV configuration directory organized by intention and trajectory, containing subdirectories such as rgb, ir, depth, depth_preview, echo, echo_groundtruth, imu, gps, as well as labels_all.txt, timestamps.csv, meta.json, and various video files. This dataset is suitable for tasks such as video classification, object detection, and multimodal intention recognition, with a total sample count ranging from 100,000 to 1,000,000.




