Explaining Image Classifiers with Multiscale Directional Image Representation

TL;DR

ShearletX is a new explanation method for image classifiers that uses shearlet transform-based sparsity constraints to produce clear, edge-focused explanations, outperforming previous smoothness-regularized methods.

Contribution

It introduces shearlet sparsity constraints for mask explanations, replacing smoothness regularization, enabling better separation of relevant fine details from nuisance patterns.

Findings

ShearletX produces more precise explanations with fewer artifacts.

It outperforms previous methods on new evaluation metrics.

Separating fine details improves interpretability of classifier decisions.

Abstract

Image classifiers are known to be difficult to interpret and therefore require explanation methods to understand their decisions. We present ShearletX, a novel mask explanation method for image classifiers based on the shearlet transform -- a multiscale directional image representation. Current mask explanation methods are regularized by smoothness constraints that protect against undesirable fine-grained explanation artifacts. However, the smoothness of a mask limits its ability to separate fine-detail patterns, that are relevant for the classifier, from nearby nuisance patterns, that do not affect the classifier. ShearletX solves this problem by avoiding smoothness regularization all together, replacing it by shearlet sparsity constraints. The resulting explanations consist of a few edges, textures, and smooth parts of the original image, that are the most relevant for the decision of the classifier. To support our method, we propose a mathematical definition for explanation artifacts and an information theoretic score to evaluate the quality of mask explanations. We demonstrate the superiority of ShearletX over previous mask based explanation methods using these new metrics, and present exemplary situations where separating fine-detail patterns allows explaining phenomena that were not explainable before.