The world's first vehicle-road collaborative autonomous driving dataset based on real-world scenarios was released on Feb 24. Now, it's available for download and use by users in China.

The world's first vehicle-road collaborative autonomous driving dataset is released on Feb 24. [Photo/kfqgw.beijing.gov.cn]

The dataset was jointly developed by the Institute for AI Industry Research (AIR) of Tsinghua University (AIR) in conjunction with the Beijing High-level Autonomous Driving Demonstration Zone, Beijing Vehicle Network Technology Development Co., Ltd., Baidu Apollo, and Beijing Academy of Artificial Intelligence. 

The dataset is the first to realize the use of 2D and 3D labeling methods from the joint perspective of the vehicle and the road within the same time and space. 

As the first open source vehicle-road collaborative autonomous driving data set in the industry and academia, it will effectively serve scientific research, the automobile industry and government agencies, effectively help all parties conduct collaborative research on vehicles and roads, municipal planning and construction, and promote the coordinated development of vehicles and roads in China.

The data in the dataset covers 10 kilometers of urban roads, 10 kilometers of highways and 28 intersections in the demonstration area, including 72,890 image frames from multiple types of sensors such as vehicle-side, road-side cameras, and vehicle-side and road-side lidars. 

Compared with datasets that only contain single-vehicle or road data, this dataset for the first time overcomes the previous problem of vehicle-road coordination in the same time and space. 

In addition, through innovations such as 3D annotation methods, the dataset will improve the 3D target detection accuracy of algorithms in the test set to reduce the amount of road-end data usage and lower communication delays and sensor usage, thereby saving costs and reducing energy consumption.