Paper - Supplementary material
This dataset is intended for the evaluation of visual-based localization and mapping systems in agriculture. It includes stereo images, IMU, GPS, and wheel encoder measurements. It was collected from a ground robot in the Illinois Autonomous Farm at the University of Illinois at Urbana-Champaign. The collection campaign took place during the Summer of 2022. Different data sequences were collected twice per week in corn fields, and less often in soybean and sorghum. This dataset exhibit high variability in terms of weather conditions and growth stages. It contains challenging features like occlusions, illumination variations, weeds, dynamic objects, and rough terrain.
This dataset is organized in different folders, classified by crop type and collection date (See Table I for further description of each folder). Within each folder, we find Rosbag (.bag) and SVO (.svo) files.
Folder Number of Time Occlusions Presence of Weather Growth-stage Rough Folder
sequences span Weeds variability variability terrain size (GB)
Cornfield1 80 4 months ✓ ✓ ✓ ✓ ✓ 584
Cornfield2 17 3 months ✓ ✓ ✓ ✓ ✓ 171
Cornfield3 2 1 week x x x x ✓ 28
Cornfield4 4 1 months ✓ x x ✓ ✓ 37
Sorghum 2 3 weeks ✓ ✓ x x ✓ 9
Soybean 12 2 weeks ✓ ✓ ✓ x ✓ 79
Sweet Corn 4 1 weeks ✓ ✓ x x ✓ 49
Others 14 3 months x x ✓ ✓ ✓ 103
If you want higher resolution images and higher depth quality than that of the rosbag files, then you can use the svo files. Most of them were recorded simultaneously with the rosbags.
cd ~/catkin_ws
source devel/setup.bash
roslaunch zed_wrapper zed2.launch svo_file:=/home/path_to_svo_file/file.svo
rostopic list
You can change the quality of the depth image and other parameters by editing the file common.yaml in the directory: catkin_ws/src/zed-ros-wrapper/zed_wrapper/params. To get the Intrinsic parameters of the camera, check topics:
In order to read the GPS and motor messages, which are user-defined ROS messages, you will need to build them in your catkin workspace. Further details.
catkin_make fpn_msgs
source devel/setup.bash
source ~/catkin_ws/devel/setup.bash
python3 extract_data_from_rosbag.py
## Sensor calibration
### Robot's coordinate frames
<div align="center">
<a href="https://github.com/jrcuaranv/terrasentia-dataset/blob/main/figures/coordinate_frames.png">
<img src="https://github.com/jrcuaranv/terrasentia-dataset/raw/main/figures/coordinate_frames.png" width="400" alt="coordinate-frames">
</a>
</div>
Download [sensor_parameters.txt](sensor_parameters.txt)
## Publication
If you find this dataset useful, please cite our paper:
Cuaran, J., Baquero Velasquez, A. E., Valverde Gasparino, M., Uppalapati, N. K., Sivakumar, A. N., Wasserman, J., ... & Chowdhary, G. (2023). Under-canopy dataset for advancing simultaneous localization and mapping in agricultural robotics. The International Journal of Robotics Research, 02783649231215372.
```bibtex
@article{cuaran2023under,
title={Under-canopy dataset for advancing simultaneous localization and mapping in agricultural robotics},
author={Cuaran, Jose and Baquero Velasquez, Andres Eduardo and Valverde Gasparino, Mateus and Uppalapati, Naveen Kumar and Sivakumar, Arun Narenthiran and Wasserman, Justin and Huzaifa, Muhammad and Adve, Sarita and Chowdhary, Girish},
journal={The International Journal of Robotics Research},
volume={43},
number={6},
pages={739--749},
year={2023},
publisher={SAGE Publications Sage UK: London, England}
}