IRRIWELL LA contains ground-level sensor recordings of apple trees in a commercial-style orchard in Potsdam, Germany, collected over three vegetation seasons (2021–2023). Each tree was recorded from both sides of the row with:

LiDAR at two wavelengths (660 nm and 905 nm), one 3D point cloud per side and wavelength;
an RGB-D camera, 20 synchronized RGB and depth frames per side.

The dataset also includes destructive leaf area references (complete defoliation of 10 trees with every leaf measured), Sentinel-2 satellite data of the orchard, weather station data, the planting scheme, and an overview spreadsheet describing every recording campaign.

Property	Value
Location	Potsdam, Germany
Period	2021–2023, 12 recording campaigns
Physical trees	142, in 14 blocks
Canopy architectures	Slender spindle, bi-tree (Bi-Baum)
Tree observations (tree × campaign)	1,284
RGB images	51,360
Depth frames	51,360
660 nm point clouds	2,568
905 nm point clouds	2,568
Total sensor files	107,856
Leaf area reference trees	10 (destructive, complete defoliation)
License	CC BY 4.0

Supported tasks

The dataset is intended for research on non-destructive canopy and plant trait estimation, for example:

tree-level leaf area estimation from LiDAR, RGB-D, or both;
segmentation of LiDAR point clouds into leaves, wood and fruit, and leaf area estimation from the segmented foliage;
prediction of LiDAR-derived canopy descriptors;
multimodal and multi-view 3D reconstruction and point cloud analysis of tree canopies;
temporal analysis of canopy development across seasons.

Repository contents

File	Content
`MRQ_YYYYMMDD.zip` (12 files)	Sensor data of one recording campaign (RGB, depth, 660 nm and 905 nm LiDAR)
`HuggingFace_Irriwell_RGBD_LiDAR.xlsx`	Overview of all campaigns: sensors, settings, archive sizes, file counts, tree IDs per block, notes
`HuggingFace_PlantingScheme.xlsx`	Orchard plot plan with tree labels, cultivars and canopy architecture
`Irriwell_LA_references.xlsx`	Leaf area references from destructive measurements
`HuggingFace_Sat_2021-2023.csv`	Vegetation indices and biophysical variables from Sentinel-2 for the orchard
`HuggingFace_SatImages_2021-2023.zip`	Sentinel-2 image stacks (GeoTIFF, 10 m and 20 m resolution)
`Irriwell_weather_hourly.csv`	Weather station data (TOSS station), 2021–2023
`4D_AppleScan.mp4`	Video of the measurement scene
`excluded_files.csv`	Files removed in v1.1, with the reason for each

The recording campaigns (dates, archives, file counts, tree IDs per block), the sensors and their settings (scan frequency, platform speed) and the notes per campaign are documented in HuggingFace_Irriwell_RGBD_LiDAR.xlsx.

Version history

Version	Files	Description
v1.1	107,856	Curated release: 294 files without a same-block front/back pair removed; updated leaf area measurements and overview spreadsheet.
v1.0-raw	108,150	Original release, available at commit `21e014367a51d62e54f59f9936e794506d1598b5`.

To download a specific version, pass the tag or commit hash as revision:

hf_hub_download("ATB-Potsdam/IRRIWELL_LA", "MRQ_20230719.zip", repo_type="dataset", revision="v1.1")

Code

The code for the paper below, including data preprocessing, LiDAR descriptor extraction, model training and leaf area estimation on this dataset, is available at:

Code: TBA

Citation

If you use this dataset, please cite:

@article{mvdescnet2026,
  title   = {Learning LiDAR-Derived Tree Geometry from RGB-D Towards Leaf Area Estimation},
  author  = {Sikouonmeu, Freddy and Zude-Sasse, Manuela and Courault, Dominique and Tsoulias, Nikolaos and Bignardi, Marco and Chanzy, Andr{\'e} and H{\"o}hne, Marina M.-C. and Atzmueller, Martin},
  journal = {TBA},
  year    = {2026}
}

Contact

Leibniz Institute for Agricultural Engineering and Bioeconomy (ATB), Potsdam, Germany.

Funding

This work was supported in part by the German Federal Ministry of Research, Technology and Space (BMFTR) under Grant 01DH21016 as part of the IRRIWELL project within the PRIMA programme (Partnership for Research and Innovation in the Mediterranean Area).

License

This dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. You may share and adapt it for any purpose, provided you give appropriate credit.

Downloads last month: 49

Total file size:

155 GB