METEOR

该数据集名为METEOR，记录了在非结构化场景中的交通模式和16种独特代理类别的多代理驾驶行为，包含了超过1000个一分钟的视频，以及超过200万帧带有边界框和GPS轨迹的标注。该数据集特别关注罕见且有趣的多代理驾驶行为，融入了天气、一天中的时间、道路状况和交通密度等因素。规模上，它包含了1250个一分钟的视频片段，以及超过200万帧的标注。该数据集的任务涉及对象检测和多代理行为预测。

Named METEOR, this dataset documents traffic patterns and multi-agent driving behaviors across 16 unique agent categories in unstructured driving scenarios. It includes over 1,000 one-minute video clips and more than 2 million annotated frames with bounding boxes and GPS trajectories. This dataset specifically focuses on rare and intriguing multi-agent driving behaviors, and incorporates contextual factors such as weather conditions, time of day, road conditions, and traffic density. In terms of scale, it comprises 1,250 one-minute video segments and over 2 million annotated frames. The tasks supported by this dataset involve object detection and multi-agent behavior prediction.