dr-pod/data-dive-campus-cams
收藏资源简介:
UIUC校园摄像头—2026年春季数据集是一个帧级存档,包含来自三个公共UIUC YouTube直播流的图像数据,捕获间隔分别为10秒(Quad)、30秒(Alma)和3秒(Morrow)。每帧图像都配有元数据,如时间戳、亮度和捕获状态。该数据集主要用于研究UIUC课程结束后行人激增如何导致校园交叉路口的公交车延误。数据集支持多种分析,包括行人密度与交通延误的相关性、雕像停留时间行为、车辆速度/轨迹估计、天气/日光信号提取等。数据集的格式为WebDataset兼容的tar文件,每摄像头每UTC小时一个文件。每个tar文件包含图像和对应的JSON元数据。数据集已知的限制包括人脸识别困难、夜间帧对比度低、部分摄像头视角固定等。
The UIUC Campus Cams — Spring 2026 dataset is a frame-level archive from three public UIUC YouTube live streams, captured at intervals of 10s (Quad), 30s (Alma), and 3s (Morrow). Each frame is paired with metadata such as timestamp, brightness, and capture status. The dataset is primarily used to study how UIUC class-release pedestrian surges cause bus delays at campus intersections. It supports various analyses, including pedestrian density vs. transit delay correlation, statue dwell-time behavior, vehicle speed/trajectory estimation, and weather/daylight signal extraction. The dataset is formatted as WebDataset-compatible tars, with one file per camera per UTC hour. Each tar contains image and JSON metadata pairs. Known limitations include difficulties in face identification, low contrast in night frames, and fixed framing for some cameras.




