Carolinas Highway Dataset (CHD)

该数据集名为CHD，它全面收录了从北卡罗来纳州和南卡罗来纳州拍摄的视频中提取的车辆高速公路轨迹，这些轨迹包含了两种不同的视角：平视和高角度。CHD数据集不仅包含了为车辆检测和追踪生成的标注，还使用了YOLOv5算法来生成边界框，并对原始轨迹数据进行了精确过滤。规模上，该数据集从16个视频中共提取了338,000条轨迹，总计1.6百万帧。其任务是车辆轨迹的检测与追踪。

The dataset is named CHD. It comprehensively collects vehicle highway trajectories extracted from videos captured in North Carolina and South Carolina, which covers two distinct perspectives: eye-level and high-angle. In addition to the annotations generated for vehicle detection and tracking, the CHD dataset also uses the YOLOv5 algorithm to generate bounding boxes and conducts precise filtering on the original trajectory data. In terms of scale, a total of 338,000 trajectories are extracted from 16 videos, totaling 1.6 million frames. The downstream task of this dataset is vehicle trajectory detection and tracking.

搜集汇总

背景与挑战

背景概述

该数据集名为CHD，专注于从北卡罗来纳州和南卡罗来纳州高速公路视频中提取的车辆轨迹，包含平视和高角度两种视角，使用YOLOv5算法生成标注并进行了数据过滤。规模上，从16个视频中提取了338,000条轨迹，总计1.6百万帧，主要用于车辆轨迹的检测与追踪任务。