Effect of Superpixel Aggregation on Explanations in LIME -- A Case Study with Biological Data

TL;DR

This study investigates how different superpixel segmentation methods affect the explanations generated by LIME for image classification, highlighting the variability in relevance areas depending on the method used.

Contribution

The paper provides an empirical comparison of three superpixel methods in LIME explanations, revealing their impact on relevance area accuracy against human references.

Findings

Quick-Shift yields the least relevance area overlap with human references.

Compact-Watershed shows the highest correspondence with human relevance areas.

Superpixel method choice significantly influences explanation quality.

Abstract

End-to-end learning with deep neural networks, such as convolutional neural networks (CNNs), has been demonstrated to be very successful for different tasks of image classification. To make decisions of black-box approaches transparent, different solutions have been proposed. LIME is an approach to explainable AI relying on segmenting images into superpixels based on the Quick-Shift algorithm. In this paper, we present an explorative study of how different superpixel methods, namely Felzenszwalb, SLIC and Compact-Watershed, impact the generated visual explanations. We compare the resulting relevance areas with the image parts marked by a human reference. Results show that image parts selected as relevant strongly vary depending on the applied method. Quick-Shift resulted in the least and Compact-Watershed in the highest correspondence with the reference relevance areas.