Street Scene 街景数据集

Street Scene 是一个用于视频异常检测的数据集。数据集包括从 USB 相机拍摄的 46 个训练和 35 个测试视频。视频包括白天俯瞰有自行车道和人行道的双车道街道场景，各种各样的活动场景，例如汽车驾驶、转弯等。该数据集共有 56847 帧用于训练，14 万 6410 帧用于测试。该数据集共包含 205 起自然发生的异常事件（乱穿马路和非法掉头等）。

Street Scene is a dataset dedicated to video anomaly detection. It includes 46 training videos and 35 testing videos captured by USB cameras. The videos feature two-lane street scenes with bike lanes and sidewalks, shot from an overhead perspective during daytime, covering various daily activities such as vehicle driving and turning. There are 56,847 frames for training and 146,410 frames for testing in total. This dataset contains 205 naturally occurring anomalous events including jaywalking and illegal U-turns.

搜集汇总

背景与挑战

背景概述

Street Scene 街景数据集是一个用于视频异常检测的数据集，包含46个训练视频和35个测试视频，共计56847帧训练数据和146410帧测试数据，记录了双车道街道场景中的205起自然发生的异常事件。数据集总大小为45.75 GB，适用于计算机视觉和目标检测等研究领域。