DADE

该数据集名为DADE，专为在动态环境中导航的自动驾驶代理在线训练和评估语义分割方法而设计。它包含了多种天气条件下，多个代理的长时间视频序列。该数据集为语义分割提供了地面真实标签，并包含了基于环境划分为单元格的场景。数据集规模达150吉字节，包含400个视频序列（其中100个为静态，300个为动态），覆盖了不同的天气状况。其任务是对动态环境中的语义分割。

This dataset, named DADE, is specifically designed for online training and evaluation of semantic segmentation methods for autonomous driving agents navigating in dynamic environments. It contains long-duration video sequences from multiple agents across various weather conditions. This dataset provides ground truth labels for semantic segmentation, and includes scenes divided into grid cells based on the surrounding environment. The dataset has a total size of 150 gigabytes, consisting of 400 video sequences (100 static and 300 dynamic) that cover diverse weather conditions. Its targeted task is semantic segmentation in dynamic environments.