ReX: Causal Responsibility EXplanations for image classifiers

---

ReX is a causal explainability tool for image classifiers. It also works on tabular and 3D data.

Given an input image and a classifier, ReX calculates a causal responsibility map across the data and identifies a minimal, sufficient, explanation.

ReX is black-box, that is, agnostic to the internal structure of the classifier. ReX finds single explanations, non-contiguous explanations (for partially obscured images), multiple independent explanations, contrastive explanations and lots of other things! It has a host of options and parameters, allowing you to fine tune it to your data.

For background information and detailed usage instructions, see our documentation.

Installation

ReX can be installed using pip. We recommend creating a virtual environment to install ReX. ReX has been tested using versions of Python >= 3.10. The following instructions assume conda:

conda create -n rex python=3.13
conda activate rex
pip install rex_xai

This should install an executable rex in your path.

Note:

By default, onnxruntime will be installed. If you wish to use a GPU, you should uninstall onnxruntime and install onnxruntime-gpu instead. You can alternatively clone the project and edit the pyproject.toml to read "onnxruntime-gpu >= 1.17.0" rather than "onnxruntime >= 1.17.0".

If you want to use ReX with 3D data, you will need to install some optional extra dependenices:

pip install 'rex_xai[3D]'

Feedback

Bug reports, questions, and suggestions for enhancements are welcome - please check the GitHub Issues to see if there is already a relevant issue, or open a new one!

How to Contribute

Your contributions are highly valued and welcomed. To get started, please review the guidelines outlined in the CONTRIBUTING.md file. We look forward to your participation!