Key features:

• Cropped RGB stickers of the rendered obstacles (for both rgb_center and rgb_highres_center cameras)
• Corresponding segmentation mask of each obstacle
• Modified point-cloud data that include the obstacle (with obstacle segmentation)
• 6 different obstacles for the following OSDaR23 sequences: 3_fire_site_3.1 - 5_station_bergedorf_5.2 - 6_station_klein_flottbek_6.1.

For licensing reasons, each sequence is released as a set of "stickers" generated with our custom UE5 simulator. The interested user must first download and extract the original OSDaR23 sequences that can be found here.

The final dataset can be generated by running the dedicated python script included in each released dataset sequence. The python script will combine original OSDaR23 images and our augmentations and produce the final dataset.

• 3_fire_site_3.1
• 5_station_bergedorf_5.2
• 6_station_klein_flottbek_6.1

OSDaR-AR was submitted for publication.
Waiting for the official proceedings, if you use OSDaR-AR in your research, please cite the following pre-print:

@misc{nesti2026osdar_ar,
  title        = {OSDaR-AR: Enhancing Railway Perception Datasets via Multi-modal Augmented Reality},
  author       = {Nesti, Federico and D'Amico, Gianluca and Marinoni, Mauro and Buttazzo, Giorgio},
  year         = {2026},
  eprint       = {2602.22920},
  archivePrefix= {arXiv},
  primaryClass = {cs.CV},
  url          = {https://arxiv.org/abs/2602.22920}
}