Custom Dataset from CARLA Simulator

该数据集旨在通过动态视觉传感器（DVS）和RGB图像捕获在各种不利天气条件下行人的过街行为。数据集分为两个子集：默认子集（晴天条件）和天气子集（包括雨、风暴和雾）。每个视频剪辑包含900帧，其中RGB分辨率为1600x600，DVS分辨率为1542x587。在规模上，默认子集包含117个晴天气候下的视频；而天气子集则包含81个不利天气条件下的视频。该数据集的任务是进行行人检测以及预测其过街行为。

This dataset is developed to capture the street-crossing behaviors of pedestrians under various adverse weather conditions using dynamic vision sensors (DVS) and RGB images. The dataset is split into two subsets: the default subset (clear-sky conditions) and the weather subset, which includes rain, storms, and fog. Each video clip consists of 900 frames, with an RGB resolution of 1600×600 and a DVS resolution of 1542×587. In terms of scale, the default subset comprises 117 video clips captured under clear-sky conditions, while the weather subset contains 81 video clips under adverse weather conditions. The tasks of this dataset are pedestrian detection and prediction of pedestrians' street-crossing behaviors.