Reproducibility Study of ”CartoonX: Cartoon Explanations of Image Classifiers”

摘要

In this reproducibility study, we verify the claims and contributions in Cartoon Explanations of Image Classifiers by Kolek et al.. These include (i) A proposed technique named CartoonX used to extract visual explanations for predictions via image classification networks, (ii) CartoonX being able to reveal piece‐wise smooth regions of the image, unlike previous methods, which extract relevant pixel‐sparse regions, and (iii) CartoonX achieving lower distortion values than these methods.