Inpainting the Gaps: A Novel Framework for Evaluating Explanation Methods in Vision Transformers

TL;DR

This paper introduces InG, a new semi-synthetic evaluation framework for explanation methods in Vision Transformers that reduces test-time distribution shift by inpainting image gaps, leading to more reliable assessments.

Contribution

The paper proposes InG, a novel inpainting-based evaluation framework for ViT explanation methods, addressing limitations of perturbation tests and providing more consistent evaluation scores.

Findings

InG improves the reliability of explanation method evaluation.

Beyond Intuition and Generic Attribution are the most consistent explanation models.

InG yields higher and more stable evaluation scores across ViT models.

Abstract

The perturbation test remains the go-to evaluation approach for explanation methods in computer vision. This evaluation method has a major drawback of test-time distribution shift due to pixel-masking that is not present in the training set. To overcome this drawback, we propose a novel evaluation framework called \textbf{Inpainting the Gaps (InG)}. Specifically, we propose inpainting parts that constitute partial or complete objects in an image. In this way, one can perform meaningful image perturbations with lower test-time distribution shifts, thereby improving the efficacy of the perturbation test. InG is applied to the PartImageNet dataset to evaluate the performance of popular explanation methods for three training strategies of the Vision Transformer (ViT). Based on this evaluation, we found Beyond Intuition and Generic Attribution to be the two most consistent explanation models. Further, and interestingly, the proposed framework results in higher and more consistent evaluation scores across all the ViT models considered in this work. To the best of our knowledge, InG is the first semi-synthetic framework for the evaluation of ViT explanation methods.