DeadLeaves

An open-source Python package for creating dead leaves images in a systematic, yet flexible way.

Core functionalities

• generating dead leaves images with properties (e.g. sizes, orientations, colors) drawn from a wide range of distributions (e.g. uniform, normal, Poisson, power-law, constant) or directly from an image.
• picking from various leaf shapes (circles, ellipsoids, rectangles, regular polygons).
• sampling in different color spaces (RGB, HSV, Gray-scale).
• applying different noise or image textures, either to the entire image or per-leaf.
• varying the image area covered by leaves, i.e. choosing between sparser or denser sampling and position mask.
• creating arbitrarily complex leaf configurations by adding dependencies between leaf features (e.g. space-dependent color gradients).

Installation

You may install deadleaves from PyPI using pip:

pip install deadleaves

OR (for developers), install from source:

1. Clone the repository from GitHub:

git clone git@github.com:ag-perception-wallis-lab/deadleaves.git

2. Install deadleaves to your local python library using pip, by running from the top-level directory:

pip install .

To install in developer/editable mode run pip install -e . at the root directory. This makes changes to files immediately usable, rather than having to reinstall the package after every change.

Dependencies

We recommend using a Python version 3.12 or newer. The dependencies should be automatically installed (at least using pip). deadleavess required dependencies are:

• PyTorch
• Torchquad
• Torchvision
• Pandas