The MuST-C dataset: The Multi-Sensor and Multi-Temporal Dataset of Multiple Crops
for In-Field Phenotyping and Monitoring
Phenotyping is crucial for understanding crop trait variation and advancing research, but is currently limited by
expensive, labor-intensive monitoring. New phenotypic trait monitoring methods are being proposed to reduce this
so-called phenotyping bottleneck via automation. These methods are often data-driven, requiring a dataset recorded
with a specific sensor and corresponding reference values for novel method development.
To this end, we present the MuST-C (Multi-Sensor, multi-Temporal, multiple Crops) dataset, which contains field
data from various sensors collected over one growing season, covering six crop species. All data are
geo-referenced
for alignment across sensors and dates. To collect our dataset, we deployed aerial and ground robotic platforms
equipped with RGB cameras, LiDARs, and multi-spectral cameras to achieve a high variety of modalities and
observations from varying viewpoints. In addition to sensor data, we also provide manually collected leaf area
index and biomass reference measurements. Our dataset enables the development of novel automatic phenotypic trait
estimation methods, allows comparisons across different sensors, and generalizability across crop species.
Quick Start
Our dataset comprises data from multiple sensors over multiple days and multiple crops, and
can be used for multiple tasks. In this example, we show how you can quickly start by performing a simple data
processing: sorting all data into plot-level data for a given plot.
- Data: Download the Sample
Data (5 GB) that contains data for a single plot in the
folder structure provided. After unzipping the sample data, you can open the data in
your
favorite software.
- Code: Get the code by cloning (or downloading the zip file of the repository) from our Git repository
https://github.com/PRBonn/MuST-C.
- Example: Our data is geo-referenced, therefore, one can extract a plot from the provided data. Inside
the
dev_kit folder you can run the following command to extract the data for plot id 198 (a sugar
beet plot):
python3 get_plot_data.py --parent_dir /path/to/dataset --output_dir path/to/output --plot 198
UAV1-RGB
UAV2-RGB
UAV2-Lidar
UAV3-RGB
UAV3-MS
UGV-RGB
UGV-Ouster
UGV-LMI
Plants
| Species |
Sugar Beet |
| Varieties |
BTS440 |
| Details |
Herbicide concentrations: 0% / 50% / 75% / 100% |
| Species |
Soy Bean |
| Varieties |
Eiko and Minngold |
| Species |
Potato |
| Varieties |
Belana and Gala |
| Species |
Maize |
| Varieties |
Khan, Popcorn Robust, Mirza, and Caramelo |
| Species |
Wheat |
| Varieties |
Granus |
| Details |
Sowing densities: 150/250/350/450 seeds m-2 |
| Species |
Wheat and Faba Beans (Intercrop) |
| Varieties |
Wheat: Sorbas, Faba: Fanfare |
Folder structure
We organized the data into super categories and split the data into platform-specific folders inside the
categories.
Extracting all provided zip files in a common directory will result in the above-shown folder structure,
which is also used by our provided code. This means the zip files contain this specific folder structure to
make the organization of the data easier.
Code
Please see our GitHub repository available at https://github.com/PRBonn/MuST-C. This repository provides
various post-processing methods that we used in the preparation of the dataset, but also more specific
examples on how to use our data. Mor specifically, we include there the following:
- Dev Kit: Python scripts to extract plot-wise data from the whole field recordings that uses th GPS
information and shape files provided with the dataset.
- Download helper scripts using wget that uses a text file to download the data packages.
- Code used for computing the LAI from scanned leaves for obtaining the destructive measurement.
- Code for obtaining the multi-spectral reflectance from the multi-spectral camera (UAV3-MS).
- Code to generate the plots used in the corresponding dataset paper.
Citation
Our corresponding article providing more details about the data generation process, the technical methodology
of the employed methods, but also further description of the platforms is currently under review and when
accepted we will provide here the full BibTeX entry for citing our work.
@article{chong2025xxx,
title = {{The MuST-C dataset: The Multi-Sensor and Multi-Temporal Dataset of Multiple Crops for In-Field Phenotyping and Monitoring}},
author = {Yue Linn Chong and Julie Kr\"amer and Erekle Chakhvashvili and Elias Marks and Felix Esser and Ansgar Dreier and
Radu Alexandru Rosu and Kevin Warstat and Ralf Pude amd Sven Behnke and Onno Muller and Uwe Rascher and Heiner Kuhlmann
and Cyrill Stachniss and Jens Behley and Lasse Klingbeil},
journal = {},
year = 2025,
note = {under review}
}
Acknowledgements
This work has been funded by been funded by the Deutsche Forschungsgemeinschaft (DFG, German Research
Foundation) under Germany's Excellence Strategy, EXC-2070 - 390732324 (PhenoRob).
The authors express their gratitude for the invaluable support and technical assistance provided by Hannah
Becker, Arnaud Kentsa, Silas Omondi, Michael Quarten, Cansu Beyaz, Svetlana Seliunina, and the team of Campus
Klein-Altendorf, especially to Magdalena Theisen, Patrick Hostnik, and Dr. Thorsten Kraska.
