IS-CAM: Integrated Score-CAM for axiomatic-based explanations

TL;DR

This paper introduces IS-CAM, an enhanced visualization method for CNNs that produces sharper attribution maps, improving interpretability and trustworthiness of these models across various architectures.

Contribution

The paper proposes IS-CAM, integrating an operation into Score-CAM to generate more precise and visually clearer attribution maps for CNN explanations.

Findings

IS-CAM produces sharper attribution maps than Score-CAM.

The method is effective across different CNN models.

Evaluations on 2000 images demonstrate versatility and robustness.

Abstract

Convolutional Neural Networks have been known as black-box models as humans cannot interpret their inner functionalities. With an attempt to make CNNs more interpretable and trustworthy, we propose IS-CAM (Integrated Score-CAM), where we introduce the integration operation within the Score-CAM pipeline to achieve visually sharper attribution maps quantitatively. Our method is evaluated on 2000 randomly selected images from the ILSVRC 2012 Validation dataset, which proves the versatility of IS-CAM to account for different models and methods.