The TUMTraf V2X Cooperative Perception Dataset is a comprehensive resource for enhancing the capabilities of autonomous vehicles and improving road safety through cooperative perception. The dataset includes 2,000 labeled point clouds and 5,000 labeled images from five roadside and four onboard sensors, covering 30k 3D boxes with track IDs and precise GPS and IMU data. It addresses challenging scenarios such as traffic violations, near-miss events, overtaking, and U-turns. The dataset is designed to provide high-quality labels through careful labeling and review processes, ensuring robustness in various weather conditions and lighting variations. To benchmark the dataset, the authors propose CoopDet3D, a cooperative multi-modal 3D object detection model. Extensive experiments demonstrate that CoopDet3D achieves a significant improvement of +14.36 3D mAP compared to a vehicle camera-LiDAR fusion model. The dataset and model are made publicly available, along with a development kit that facilitates data loading, annotation, and evaluation. The paper also discusses the sensor setup, calibration, and registration processes, ensuring accurate and synchronized data from both onboard and roadside sensors. The dataset is structured to include diverse traffic scenarios, and the development kit supports various functionalities such as data augmentation, visualization, and model evaluation. Finally, the authors highlight the importance of V2X datasets in overcoming limitations of single-viewpoint datasets, such as occlusions and limited field of view. The proposed dataset and model are expected to advance the field of cooperative perception and enhance the safety and efficiency of autonomous vehicles.The TUMTraf V2X Cooperative Perception Dataset is a comprehensive resource for enhancing the capabilities of autonomous vehicles and improving road safety through cooperative perception. The dataset includes 2,000 labeled point clouds and 5,000 labeled images from five roadside and four onboard sensors, covering 30k 3D boxes with track IDs and precise GPS and IMU data. It addresses challenging scenarios such as traffic violations, near-miss events, overtaking, and U-turns. The dataset is designed to provide high-quality labels through careful labeling and review processes, ensuring robustness in various weather conditions and lighting variations. To benchmark the dataset, the authors propose CoopDet3D, a cooperative multi-modal 3D object detection model. Extensive experiments demonstrate that CoopDet3D achieves a significant improvement of +14.36 3D mAP compared to a vehicle camera-LiDAR fusion model. The dataset and model are made publicly available, along with a development kit that facilitates data loading, annotation, and evaluation. The paper also discusses the sensor setup, calibration, and registration processes, ensuring accurate and synchronized data from both onboard and roadside sensors. The dataset is structured to include diverse traffic scenarios, and the development kit supports various functionalities such as data augmentation, visualization, and model evaluation. Finally, the authors highlight the importance of V2X datasets in overcoming limitations of single-viewpoint datasets, such as occlusions and limited field of view. The proposed dataset and model are expected to advance the field of cooperative perception and enhance the safety and efficiency of autonomous vehicles.