运动场景下的大规模多目标跟踪数据集 (MOT Windsurfing2025-5TASK)

该数据集基于运动相机拍摄的帆板运动的动态场景的视频数据，对其中的目标（人）进行人工逐帧标注生成。以两种数据格式提供：含标注框的视频MP4格式和含目标跟踪标注结果的标准VOC格式文件XML格式。 本数据集为自制数据集，原始视频影像数据为自采集，标注数据集使用开源标注工具CAVT网页版生成。总数据量 106M，包含5个独立的视频跟踪场景。<br> The dataset is generated by manual frame-by-frame annotation of targets (humans) in video data of dynamic windsurfing scenes captured by a sports camera. This is a self-constructed dataset. The raw video data was independently collected, and the annotated dataset was generated using the open-source web-based CVAT tool. <br>This dataset is generated by manual frame-by-frame annotation of targets (humans) in video data of dynamic windsurfing scenes captured by a sports camera. It is provided in two data formats: video in MP4 format with bounding boxes, and XML files in standard VOC format containing object tracking annotation results. <br>This is a self-constructed dataset; the raw video image data was independently collected, and the annotated dataset was generated using the open-source web-based CVAT tool. The total data size is 106MB, including 5 independent video tracking scenarios. <br>The dataset is generated by manual frame-by-frame annotation of targets (humans) in video data of dynamic windsurfing scenes captured by a sports camera. This is a self-constructed dataset. The raw video data was independently collected, and the annotated dataset was generated using the open-source web-based CVAT tool. <br>

搜集汇总

背景与挑战

背景概述

该数据集是一个针对帆板运动场景的大规模多目标跟踪数据集，包含5个独立视频场景，总数据量105.56 MB，基于大疆运动相机于2025年6月在福建漳州采集，并通过人工逐帧标注生成MP4和XML格式文件。其特点在于专注于动态运动环境下的目标（人）跟踪，属于地理信息系统学科，适用于运动分析和计算机视觉研究，数据质量经过严格审查，可靠性高。