Open-Canopy: Towards Very High Resolution Forest Monitoring

 

This is the official repository associated with the pre-print: "Open-Canopy: Towards Very High Resolution Forest Monitoring".

This repository includes the code needed to reproduce all experiments in the paper.

• Datapaper : Pre-print on arXiv: https://arxiv.org/abs/2407.09392.

• Code : https://github.com/fajwel/Open-Canopy

• Dataset link : https://huggingface.co/datasets/AI4Forest/Open-Canopy.

• Size : Approximately 360GB, including predictions on test set and pretrained models.

Context & Data

Estimating canopy height and canopy height change at meter resolution from satellite imagery has numerous applications, such as monitoring forest health, logging activities, wood resources, and carbon stocks. However, many existing forestry datasets rely on commercial or closed data sources, restricting the reproducibility and evaluation of new approaches. To address this gap, we introduce Open-Canopy, an open-access and country-scale benchmark for very high resolution (1.5 m) canopy height estimation. Covering more than 87,000 km2 across France, Open-Canopy combines SPOT 6-7 satellite imagery with high resolution aerial LiDAR data. Additionally, we propose a benchmark for canopy height change detection between two images taken at different years, a particularly challenging task even for recent models. To establish a robust foundation for these benchmarks, we evaluate a comprehensive list of state-of-the-art computer vision models for canopy height estimation.

Examples of canopy height estimation

Example of canopy height change estimation

Dataset Structure

A full description of the dataset can be found in the supplementary material of the Open-Canopy article.

Our training, validation, and test sets cover most of the French territory. Test tiles are separated from train and validation tiles by a 1km buffer (a).

For each tile, we provide VHR images at a 1.5 m resolution (b) and associated LiDAR-derived canopy height maps (c).

Installation & Usage

See the Open-Canopy GitHub. Note: in the first version of the dataset, non classified points were not taken into account in order to compute canopy height models (CHM) from LiDAR point clouds. The affected pixels can be masked using the provided lidar classification rasters (class 1). Starting January 2025, we also provide a second version of the CHMs where non classified points are included for computations (folder lidar_v2). This can lead to slighlty better metrics (gain about 0.05m on height MAE for the best model), although it affects less than 0.5% of pixels. However, use the first version of the CHMs to reproduce results of the paper. IGN is also starting to release pre-computed CHMs in some areas. When and where available, we recommend to use the CHMs released by IGN.

Pretrained models

Unet and PVTv2 models trained on Open-Canopy are available in the pretrained_models folder of the dataset.

Reference

Please include a citation to the following article if you use the Open-Canopy dataset:

@article{fogel2024opencanopy,
      title={Open-Canopy: A Country-Scale Benchmark for Canopy Height Estimation at Very High Resolution},
      author={Fajwel Fogel and Yohann Perron and Nikola Besic and Laurent Saint-André and Agnès Pellissier-Tanon and Martin Schwartz and Thomas Boudras and Ibrahim Fayad and Alexandre d'Aspremont and Loic Landrieu and Philippe Ciais},
      year={2024},
      eprint={2407.09392},
      publisher = {arXiv},
      url={https://arxiv.org/abs/2407.09392},
}

Acknowledgements

This paper is part of the project AI4Forest, which is funded by the French National Research Agency (ANR), the German Aerospace Center (DLR) and the German federal ministry for education and research (BMBF).

The experiments conducted in this study were performed using HPC/AI resources provided by GENCI-IDRIS (Grant 2023-AD010114718 and 2023-AD011014781) and Inria.

Dataset license

The "OPEN LICENCE 2.0/LICENCE OUVERTE" is a license created by the French government specifically for the purpose of facilitating the dissemination of open data by public administration. If you are looking for an English version of this license, you can find it at the official github page.

As stated by the license :

• Applicable legislation: This licence is governed by French law.
• Compatibility of this licence: This licence has been designed to be compatible with any free licence that at least requires an acknowledgement of authorship, and specifically with the previous version of this licence as well as with the following licences: United Kingdom’s “Open Government Licence” (OGL), Creative Commons’ “Creative Commons Attribution” (CC-BY) and Open Knowledge Foundation’s “Open Data Commons Attribution” (ODC-BY).

Authors

Fajwel Fogel (ENS), Yohann Perron (LIGM, ENPC, CNRS, UGE, EFEO), Nikola Besic (LIF, IGN, ENSG), Laurent Saint-André (INRAE, BEF), Agnès Pellissier-Tanon (LSCE/IPSL, CEA-CNRS-UVSQ), Martin Schwartz (LSCE/IPSL, CEA-CNRS-UVSQ), Thomas Boudras (LSCE/IPSL, CEA-CNRS-UVSQ), Ibrahim Fayad (LSCE/IPSL, CEA-CNRS-UVSQ, Kayrros), Alexandre d'Aspremont (CNRS, ENS, Kayrros), Loic Landrieu (LIGM, ENPC, CNRS, UGE), Philippe Ciais (LSCE/IPSL, CEA-CNRS-UVSQ).