DeepTrees🌳

Tree Crown Segmentation and Analysis in Remote Sensing Imagery with PyTorch

Authors:

Taimur Khan (UFZ) Caroline Arnold (Hereon) Harsh Grover (Hereon)

Version:

1.6.1

Contents:

DeepTrees is a PyTorch-based library for end-to-end tree crown segmentation and analysis in multispectral remote sensing imagery.

Highlights:

  • User-friendly, flexible, and GPU-optimized for efficiency.

  • Supports 4-channel imagery (RGBi) with built-in PyTorch data loaders.

  • Simple API for training and evaluating U-Net-based tree segmentation and distance transform models.

  • Generates pixel-entropy maps for active learning and fine-tuning.

  • Computes vegetation indices and allometric metrics.

  • Pre-trained models and sample labeled datasets for DOP imagery in Central Germany.

  • Easily configurable via a YAML file.

Future Work:

  • Integration of Geospatial Foundation Model (GeoFM) backbones.

  • Tree species classification and height estimation models.

  • Integration of 3D point cloud data for tree crown volume estimation.

  • Tree vitality composites derived from Land Surface Temperature (LST) and multispectral vegetation indices.

DeepTrees is a result of the DeepTrees project, a collaboration between the Helmholtz Center for Environmental Research – UFZ and the Helmholtz AI initiative.

Installation

From PyPi registry:

pip install deeptrees

or from source:

git clone https://codebase.helmholtz.cloud/taimur.khan/DeepTrees.git
python3 setup.py install

Cite As

If you use DeepTrees in your research, please cite it as follows:

APA Citation

Khan, T., Arnold, C., & Grover, H. (2025). DeepTrees: Tree crown segmentation and analysis in remote sensing imagery with PyTorch. arXiv. https://doi.org/10.48550/arXiv.XXXXX.YYYYY

BibTeX Citation

@article{khan2025deeptrees,
  author    = {Taimur Khan and Caroline Arnold and Harsh Grover},
  title     = {DeepTrees: Tree Crown Segmentation and Analysis in Remote Sensing Imagery with PyTorch},
  year      = {2025},
  journal   = {ResearchGate},
  archivePrefix = {ResearchGate},
  eprint    = {10.13140/RG.2.2.32837.36329},
  doi    = {http://dx.doi.org/10.13140/RG.2.2.32837.36329},
  primaryClass = {cs.CV}
}

License

This package is license under the MIT License. See the LICENSE file for details.