2021年清华大学AIR研究院采集的大规模真实场景车路协同运动数据集DAIR-V2X-Seq

清华大学智能产业研究院(AIR)依托北京市高级别自动驾驶示范区，继推出全球首个基于真实自动驾驶场景的车路协同3D目标检测数据集DAIR-V2X后，为了加速研究车路协同时序感知和轨迹预测，发布全球首个基于真实道路场景的大规模时序车路协同数据集V2X-Seq，进一步探索车路协同自动驾驶的落地模式。 数据集分为路端数据和车端数据，分别由路端摄像头和车端摄像头/激光雷达所采集，除了包含xy坐标，车辆种类，车辆id，车辆速度，车辆属性等常规数据外，还包含当前道路的交通灯信息，车辆长度、宽度等额外信息。Tsinghua University Institute for AI Industry Research (AIR), leveraging the Beijing High-level Autonomous Driving Demonstration Zone, first launched DAIR-V2X, the world's first real-world autonomous driving scenario-based vehicle-road collaborative 3D object detection dataset. To accelerate research on temporal perception and trajectory prediction for vehicle-road collaboration and further explore the deployment models of vehicle-road collaborative autonomous driving, the institute has now released V2X-Seq, the world's first large-scale temporal vehicle-road collaborative dataset based on real road scenarios. The dataset comprises two subsets: roadside data and on-board data, collected by roadside cameras and on-board cameras/LiDAR respectively. Beyond conventional data such as xy coordinates, vehicle category, vehicle ID, vehicle speed and vehicle attributes, it also includes additional information including traffic light conditions of the current road, vehicle length and width.

搜集汇总

背景与挑战

背景概述

该数据集由清华大学智能产业研究院(AIR)在2021年采集，是全球首个基于真实道路场景的大规模时序车路协同数据集，旨在支持车路协同时序感知和轨迹预测研究。它包含路端和车端数据，涵盖坐标、车辆信息、交通灯等丰富属性，总数据量为31.21GB。