DeepTree

Deep learning approaches for tree species estimation from remote sensing data.

This repository contains code, and workflows for using airborne and satellite remote sensing data.

Repository Structure

The repository is organized into four main sections, each corresponding to a specific study:

├── ALSComposition/
├── FusionComposition/
├── LLPEstimator/
├── TreeEstimator/

Each folder includes code, data processing pipelines, and model implementations used in the associated work.

Studies

• ALSComposition - Estimating tree species composition from airborne laser scanning data using point-based deep learning models

  → Description: Estimation of tree species composition from ALS point clouds using point-based deep learning, combining PointAugment GAN and DGCNN models.

  → Publication: https://doi.org/10.1016/j.isprsjprs.2023.12.008

• FusionComposition - Tree species proportion prediction using airborne laser scanning and Sentinel-2 data within a deep learning based dual-stream data fusion approach

  → Description: Dual-stream deep learning approach for fusing ALS point clouds and multitemporal Sentinel-2 imagery to estimate tree species proportions.

  → Publication: https://doi.org/10.1080/01431161.2025.2521072

• LLPEstimator - Using weakly supervised deep learning to derive individual tree species and plot-level species composition from airborne laser scanning data

  → Description: Weakly supervised deep learning approach for individual tree species classification and plot-level composition estimation from ALS data.

  → Publication:

• TreeEstimator - Individual tree species prediction using airborne laser scanning data and derived point-cloud metrics within a dual-stream deep learning approach

  → Description: Dual-stream deep learning framework for individual tree species classification using ALS point clouds and point-cloud metrics.

  → Publication: https://doi.org/10.1016/j.jag.2025.104877

Data

Datasets are not included due to size and/or licensing constraints.

Citation

If you use this repository, please cite the corresponding publications:

ALSComposition

@article{Murray2024EstimatingTS,
  title={Estimating tree species composition from airborne laser scanning data using point-based deep learning models},
  author={Brent Murray and Nicholas C. Coops and Lukas Winiwarter and Joanne C. White and Adam Dick and Ignacio Barbeito and Ahmed Ragab},
  journal={ISPRS Journal of Photogrammetry and Remote Sensing},
  year={2024},
  url={https://api.semanticscholar.org/CorpusID:266711000},
  doi={https://doi.org/10.1016/j.isprsjprs.2023.12.008}
}

FusionComposition

@article{Murray2025TreeSP,
  title={Tree species proportion prediction using airborne laser scanning and Sentinel-2 data within a deep learning based dual-stream data fusion approach},
  author={Brent Murray and Nicholas C. Coops and Joanne C. White and Adam Dick and Ahmed Ragab},
  journal={International Journal of Remote Sensing},
  year={2025},
  volume={46},
  pages={5436 - 5464},
  url={https://api.semanticscholar.org/CorpusID:279559010},
  doi={https://doi.org/10.1080/01431161.2025.2521072}
}

LLPEstimator

Will update upon acceptance of publication.

TreeEstimator

@article{Murray2025IndividualTS,
  title={Individual tree species prediction using airborne laser scanning data and derived point-cloud metrics within a dual-stream deep learning approach},
  author={Brent Murray and Nicholas C. Coops and Joanne C. White and Adam Dick and Ignacio Barbeito and Ahmed Ragab},
  journal={Int. J. Appl. Earth Obs. Geoinformation},
  year={2025},
  volume={144},
  pages={104877},
  url={https://api.semanticscholar.org/CorpusID:281619636},
  doi={https://doi.org/10.1016/j.jag.2025.104877}
}

Notes

This repository is actively maintained and may evolve as additional experiments and studies are completed.