full-scale-leaves

The repository contains scripts for creating and visualizing leaf venation networks.

Table of Contents

• Setup
• Usage
• License

Setup

1. Install Python

The project requires Python, and has been successfully tested on v. 3.12.7. It can be downloaded from python.org.

2. Clone the repository

git clone https://github.com/kirkegaardlab/full-scale-leaves.git
cd full-scale-leaves

3. Make a virtual environment

python -m venv env
source env/bin/activate

4. Install the required packages

This project uses jax, which comes in different versions depending on, e.g., hardware, OS, etc. Please see https://jax.readthedocs.io/en/latest/installation.html for installation instructions. The other dependencies can be installed in the usual way:

pip install -r requirements.txt

5. Download and extract the raw image data into the data directory

The data provided here consists both of the original leaf images, as well as the segmentation probabilities.

wget -O data/images_originals.zip https://zenodo.org/records/13991723/files/images_originals.zip
wget -O data/probabilities.zip https://zenodo.org/records/13991723/files/probabilities.zip

6. Extract the files

unzip -d data/ data/images_originals.zip
unzip -d data/ data/probabilities.zip

Usage

In the following, we show the whole pipeline using the leaf image file IMG_7627.tiff as an example.

Preprocessing

In order to run the model, we first need to extract the network from the segmentation probabilities, reduce the network size by pruning, and add (possibly zero) edges to the network:

cd scripts
python extract.py --im=IMG_7627
python reduce.py --im=IMG_7627
python add.py --im=IMG_7627

Model

To run a simulation on the preprocessed data, do:

python scripts/main.py --im=IMG_7627

The model output will be stored in data/model_output.

The three most important flags to consider are:

• --im, which sets the specific image data used in the script. If no flag is given, an artificial network with a regular grid is used instead.
• --sf, which sets the sink fluctuation amplitude (default = 0.0).
• --gpu, which allows the use of a GPU, if it is available to the user, and if the libraries and dependencies have been installed correctly.

Additionally, to run on a full network instead of running on 10 individual clusters, set:

• --full, which is particularly useful for artificial networks (as they are usually much smaller than full-scale leaf networks).

For information on all flags, consult the module src/param_parser.py.

Visualization

To visualize the model output, different scripts are provided for different purposes. Note that the same flags must be used as the ones used when generating the corresponding data.

• scripts/analysis/comparison.py, which compares the input and output networks.
• scripts/analysis/artificial.py, which simply plots the output artificial leaf network.
• scripts/analysis/flows.py, which plots the input network with the average resulting flow directions.

License

This project is licensed under the MIT License - see the LICENSE file for details.