GIFT: A Framework for Global Interpretable Faithful Textual Explanations of Vision Classifiers

Introduction

This is the official repository for the paper:

GIFT: A Framework for Global Interpretable Faithful Textual Explanations of Vision Classifiers,
Éloi Zablocki, Valentin Gerard, Amaia Cardiel, Eric Gaussier, Matthieu Cord, Eduardo Valle

GIFT is a framework for generating global, interpretable textual explanations of vision classifiers, combining counterfactual visual explanations with VLMs and LLMs.

Installation and preparation

coming soon