https://github.com/martibosch/swiss-urban-trees

Draft for urban tree detection in Swizterland
https://github.com/martibosch/swiss-urban-trees

Last synced: about 2 months ago
JSON representation

Draft for urban tree detection in Swizterland

• Host: GitHub
• URL: https://github.com/martibosch/swiss-urban-trees
• Owner: martibosch
• License: gpl-3.0
• Created: 2025-01-15T07:41:06.000Z (3 months ago)
• Default Branch: main
• Last Pushed: 2025-03-03T21:23:29.000Z (about 2 months ago)
• Last Synced: 2025-03-03T22:26:15.469Z (about 2 months ago)
• Language: Jupyter Notebook
• Size: 9.3 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
[![pre-commit.ci status](https://results.pre-commit.ci/badge/github/martibosch/swiss-urban-trees/main.svg)](https://results.pre-commit.ci/latest/github/martibosch/swiss-urban-trees/main)

[![GitHub license](https://img.shields.io/github/license/martibosch/swiss-urban-trees.svg)](https://github.com/martibosch/swiss-urban-trees/blob/main/LICENSE)

# Swiss urban tree inventory

Urban tree crown detection and species identification from aerial imagery in Switzerland using deep learning.

![Example results of the crop model for different patch sizes](reports/figures/crop-model-patch-size.png)

*This repository is a draft/work in process, expect many changes/updates soon*.

## Example notebooks

- [Train/test split](https://github.com/martibosch/swiss-urban-trees/blob/main/notebooks/train-test-split.ipynb): using image descriptors and potentially tree inventories to select initial training tiles.

- [Train/fine-tune a model to predict tree crowns](https://github.com/martibosch/swiss-urban-trees/blob/main/notebooks/train-crown.ipynb): evaluate results of the [DeepForest](https://deepforest.readthedocs.io/en/latest) pre-trained model and fine-tune it with our annotations.

- [Multi species model](https://github.com/martibosch/swiss-urban-trees/blob/main/notebooks/multi-species-model.ipynb): train a model to predict tree species from the detected crowns.

- [LiDAR to tree crowns](https://github.com/martibosch/swiss-urban-trees/blob/main/notebooks/lidar-to-crown.ipynb): explore the use of LiDAR data to predict tree crowns/species to assist manual annotations or reduce the need for them.

## Instructions to reproduce

1. Create a conda environment:

```bash

snakemake -c1 create_environment

```

2. Activate it (if using conda, replace `mamba` for `conda`):

```bash

mamba activate swiss-urban-trees

```

3. Register the IPython kernel for Jupyter:

```bash

snakemake -c1 register_ipykernel

```

## Acknowledgments

- Based on the [cookiecutter-data-snake :snake:](https://github.com/martibosch/cookiecutter-data-snake) template for reproducible data science.