Tree crown delineation in a sample drone RGB image.
The delineation of individual trees in remote sensing images is an key task in forest analysis. As part of Sebastian Hickman's AI4ER MRes project, titled 'Detecting changes in tall tree height with machine learning, LiDAR, and RGB imagery', the authors propose the detectreeRGB model, an implementation of Mask R-CNN from Detectron2 to perform tree crown delineation from RGB imagery.
In this notebook, we demonstrate how scivision can assist in discovering a pretrained detectreeRGB model provided by Hickman et al (2021), and then use it to delineate crowns from a sample drone RGB image dataset.
Further details of the detectreeRGB can be found in the original model repository.
The notebook is designed to be launched from Binder.
- Click the Launch Binder button at the top level of the repository
You may also download the notebook from GitHub to run it locally:
- Open your terminal
- Check your conda install with
conda --version. If you don't have conda, install it by following these instructions (see here) - Clone the repository into your current folder
git clone https://github.com/scivision-gallery/tree-crown-detection - Move into the cloned repository,
cd tree-crown-detection - Install the dependencies in a new environment
conda env create -f environment.yml - Activate the installed environment,
conda activate treecrown-detection_detectreeRGB - Launch the jupyter interface of your preference, notebook,
jupyter notebookor labjupyter lab
A full conda environment can be found in environment_full.yml. If you are having issues creating the conda environment, this file contains details of all the packages and their versions which are required to run these notebooks, including secondary and tertiary dependencies.
This notebook was supported by the outcomes of Sebastian Hickman's AI4ER MRes project. The scivision team thanks the individuals and institutions involved in the project, in particular Sebastian for providing one of the trained models and sample images used in this notebook.