Integrated Grad-CAM: Sensitivity-Aware Visual Explanation of Deep Convolutional Networks via Integrated Gradient-Based Scoring

TL;DR

This paper proposes Integrated Grad-CAM, a sensitivity-aware visualization method that improves the interpretability of CNNs by accurately measuring feature importance through gradient path integrals, enhancing object localization and model understanding.

Contribution

It introduces a novel gradient path integral approach to improve Grad-CAM's accuracy in visual explanations of CNNs, addressing underestimation issues in feature importance measurement.

Findings

Enhanced accuracy in feature importance estimation

Improved object localization performance

Better model interpretability

Abstract

Visualizing the features captured by Convolutional Neural Networks (CNNs) is one of the conventional approaches to interpret the predictions made by these models in numerous image recognition applications. Grad-CAM is a popular solution that provides such a visualization by combining the activation maps obtained from the model. However, the average gradient-based terms deployed in this method underestimates the contribution of the representations discovered by the model to its predictions. Addressing this problem, we introduce a solution to tackle this issue by computing the path integral of the gradient-based terms in Grad-CAM. We conduct a thorough analysis to demonstrate the improvement achieved by our method in measuring the importance of the extracted representations for the CNN's predictions, which yields to our method's administration in object localization and model interpretation.