SimBEV2X: A Large-Scale Dataset and Data Generation Tool for Multi-Task Vehicle-to-Everything Cooperative Perception

Goodar Mehr, Sepideh Gohari, Montasir Abbas, Azim Eskandarian
Virginia Commonwealth University (VCU), Virginia Tech
Dataset Paper Code arXiv


SimBEV2X is an advanced vehicle-to-everything synthetic data generation tool built on the CARLA simulator. SimBEV2X automatically creates highly randomized driving scenarios to collect rich multi-modal sensor data alongside various types of ground truth, including 3D object bounding boxes with unique track IDs, HD map information, BEV segmentation maps, and 3D semantic occupancy voxel grids from both vehicles and road-side units (RSUs).

The SimBEV2X Dataset

The SimBEV2X dataset is the largest V2X perception dataset to date. The dataset comprises 258 scenes, each involving up to 8 connected vehicles and up to 4 RSUs across a variety of road networks. The SimBEV2X Dataset is an order of magnitude larger than existing V2X datasets and contains 102,200 frames, 588,520 lidar point clouds, more than 3 million images, over 27 million object bounding boxes, and a comprehensive set of BEV segmentation maps and 3D semantic occupancy voxel grids. You can download the dataset from here. See here for more information about the dataset.

Unless specifically labeled otherwise, the SimBEV2X Dataset is provided to you under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Public License (“CC BY-NC-SA 4.0”). The CC BY-NC-SA 4.0 may be accessed here. When you download or use the SimBEV2X Dataset from our websites or elsewhere, you are agreeing to comply with the terms of CC BY-NC-SA 4.0 as applicable.

BibTeX

@article{mehr2026simbev2x,
  title={SimBEV2X: A Large-Scale Dataset and Data Generation Tool for Multi-Task Vehicle-to-Everything Cooperative Perception},
  author={Mehr, Goodarz and Gohari, Sepideh and Abbas, Montasir and Eskandarian, Azim},
  journal={arXiv preprint arXiv:2607.23910},
  year={2026}
}