thisistaimur/DeepTrees_Halle

DeepTrees Halle DOP20 labels + imagery

This is a sample dataset for model training and fine-tuning in tree crown segmentation tasks using the DeepTrees Python package.

Overview of subtiles with sample of labels:

Dataset Details

We have taken a single Multispectral (RGBi) 2x2 km DOP20 image tile for Halle, Sachsen-Anhalt, from LVermGeo ST for the year of 2022.

TileID from source: 32_704_5708_2

We then sliced the tiles into subtiles of 100x100m, resulting in 400 subtiles. These are provided as 4-band raster .tif files.

We are manually labelling tree crowns in these subtiles based on an entropy-based active learning approach. The label classes have been provided below.

The labeling is done in QGIS and the polygon vectors are provided as ESRI-shape .shp files per subtile.

Label Classes

0 = tree

1 = cluster of trees

2 = unsure

3 = dead trees (haven’t added yet)

Cite the dataset

@misc {taimur_khan_2025,
    author       = { {Taimur Khan} },
    title        = { DeepTrees_Halle (Revision 0c528b9) },
    year         = 2025,
    url          = { https://huggingface.co./datasets/thisistaimur/DeepTrees_Halle },
    doi          = { 10.57967/hf/4213 },
    publisher    = { Hugging Face }
}

License

This repository is made avaiable under The Prosperity Public License 3.0.0. A copy of the license can be found in the LICENSE.md file.

References

• Halle DOP20: © GeoBasis-DE / LVermGeo ST 2022