Traffic4Cast 2021 Dataset

该数据集包含了来自四个城市（柏林、伊斯坦布尔、墨尔本、芝加哥）的交通状态数据，以网格地图的形式呈现。每个网格单元汇总了大约100米×100米区域内的GPS测量数据，数据记录间隔为5分钟。此外，该数据集包含了2019年6个月的标记数据，这些数据被分为4个折叠用于训练和验证。动态交通状态在60分钟的时间窗口内编码，并对各种未来时间点的交通状态进行预测。网格地图的规模为495×436像素，包含8个通道，分别代表交通量和速度。该数据集的任务是交通状态预测。

This dataset contains traffic state data from four cities, namely Berlin, Istanbul, Melbourne, and Chicago, presented in the format of grid maps. Each grid cell aggregates GPS measurement data within an approximately 100m × 100m area, with a data recording interval of 5 minutes. Additionally, the dataset includes labeled data spanning 6 months in 2019, which is split into 4 folds for training and validation. Dynamic traffic states are encoded over a 60-minute time window, and traffic states at various future time points are predicted. The grid maps have a size of 495 × 436 pixels and contain 8 channels that respectively represent traffic volume and speed. The core task of this dataset is traffic state prediction.