CoolFace
Datasetpublic

cifasis-conicet/rosariov2

The Rosario Dataset v2 Multi-modal dataset for agricultural robotics, recorded with a ground robot navigating soybean fields near Rosario, Argentina. It includes stereo and RGB-D imagery, IMU, wheel odometry, and both conventional and post-processed (PPK) GNSS, together with a post-processed pose ground-truth, across 6 recorded sequences. This repository mirrors the dataset's official distribution point so it remains available if the primary server is unreachable. Website:… See the full description on the dataset page: https://huggingface.co/datasets/cifasis-conicet/rosariov2.

sourceHugging Facecc-by-4.0updated 10d agoView on Hugging Face
1likes419downloads
Dataset Card

The Rosario Dataset v2

Multi-modal dataset for agricultural robotics, recorded with a ground robot navigating soybean fields near Rosario, Argentina. It includes stereo and RGB-D imagery, IMU, wheel odometry, and both conventional and post-processed (PPK) GNSS, together with a post-processed pose ground-truth, across 6 recorded sequences.

This repository mirrors the dataset's official distribution point so it remains available if the primary server is unreachable.

  • —Website: https://cifasis.github.io/rosariov2/
  • —Code & tools: https://github.com/CIFASIS/rosariov2 (extraction, conversion, calibration and replay scripts, Docker environments, and full documentation)

Citation

Please cite our work if you use this dataset:

text
@article{doi:10.1177/02783649251368909,
  author = {Nicolás Soncini and Javier Cremona and Erica Vidal and Maximiliano García and Gastón Castro and Taihú Pire},
  title ={The Rosario dataset v2: Multi-modal dataset for agricultural robotics},
  journal = {The International Journal of Robotics Research},
  year = {2025},
  pages = {02783649251368909},
  doi = {10.1177/02783649251368909},
  URL = {https://doi.org/10.1177/02783649251368909}
}

Dataset structure

  • —sequences/: the 6 recorded sessions, as ROS1 rosbags, plus add-on rosbags per sequence:
  • —<sequence_id>.compressed.bag: base rosbag (stereo/RGB-D imagery, IMU, wheel odometry).
  • —<sequence_id>_conventional_gnss.bag: conventional (non post-processed) GNSS data.
  • —<sequence_id>_ppk_gnss.bag: Post-Processed Kinematic (PPK) GNSS data.
  • —<sequence_id>_pgt.bag: post-processed pose ground-truth.
  • —<sequence_id>_odom_fix.bag: corrected wheel odometry (see the code repo's docs/odometry_fix.md), published as /odom_fix, meant to be merged with the base rosbag.
  • —sequences_plain/: select topics extracted from the rosbags as plaintext (CSV/TUM), with its own README.
  • —calibration/: calibration recordings (IMU bias, camera/IMU intrinsics and extrinsics) and the resulting calibration files.
  • —raw_gnss/: raw GNSS receiver logs as produced during data collection.
  • —supplementary_dataset/: an additional dataset recorded on a separate date at the same premises, with its own README and CHANGELOG.

Sequences

#Sequence IDDurationDistanceDate
12023-12-22-13-14-16940s777m2023-12-22
22023-12-22-14-29-431011s904m2023-12-22
32023-12-22-16-31-08943s950m2023-12-22
42023-12-26-13-39-432506s2254m2023-12-26
52023-12-26-15-10-15796s703m2023-12-26
62023-12-26-15-48-381862s1744m2023-12-26

License

All data in this dataset is licensed under a Creative Commons 4.0 Attribution License (CC BY 4.0). Please check the LICENSE file included in this repository for full terms.

Changelog

Changes to the dataset itself are tracked in CHANGELOG.md.