The paper introduces RCooper, a real-world, large-scale dataset for roadside cooperative perception, aiming to enhance the capabilities of autonomous driving and traffic management. RCooper addresses the limitations of single-infrastructure sensor systems by enabling comprehensive understanding of traffic areas through cross-infrastructure cooperation. The dataset includes 50k images and 30k point clouds, covering two typical traffic scenes: intersections and corridors. It is manually annotated with 3D bounding boxes and trajectories for ten semantic classes. The paper also presents two cooperative perception tasks—3D object detection and tracking—and reports comprehensive benchmarks with state-of-the-art methods. The results demonstrate the effectiveness of roadside cooperation and highlight the need for further research in areas such as data heterogeneity, cooperative representation, and perception performance. The dataset and benchmarks are available at <https://github.com/AIR-THU/DAIR-RCooper>.The paper introduces RCooper, a real-world, large-scale dataset for roadside cooperative perception, aiming to enhance the capabilities of autonomous driving and traffic management. RCooper addresses the limitations of single-infrastructure sensor systems by enabling comprehensive understanding of traffic areas through cross-infrastructure cooperation. The dataset includes 50k images and 30k point clouds, covering two typical traffic scenes: intersections and corridors. It is manually annotated with 3D bounding boxes and trajectories for ten semantic classes. The paper also presents two cooperative perception tasks—3D object detection and tracking—and reports comprehensive benchmarks with state-of-the-art methods. The results demonstrate the effectiveness of roadside cooperation and highlight the need for further research in areas such as data heterogeneity, cooperative representation, and perception performance. The dataset and benchmarks are available at <https://github.com/AIR-THU/DAIR-RCooper>.