MetaCAM: Ensemble-Based Class Activation Map

TL;DR

MetaCAM is an ensemble approach that combines multiple CAM methods to produce more accurate and trustworthy visual explanations for CNN predictions, especially in high-stakes fields like medicine.

Contribution

It introduces MetaCAM, an ensemble-based method using consensus among CAMs, along with a new summarization technique (CRE) and adaptive thresholding to enhance explanation quality.

Findings

MetaCAM outperforms individual CAMs in explanation accuracy.

Adaptive thresholding improves CAM performance on pixel perturbation tests.

CRE effectively summarizes large-scale ensemble experiments.

Abstract

The need for clear, trustworthy explanations of deep learning model predictions is essential for high-criticality fields, such as medicine and biometric identification. Class Activation Maps (CAMs) are an increasingly popular category of visual explanation methods for Convolutional Neural Networks (CNNs). However, the performance of individual CAMs depends largely on experimental parameters such as the selected image, target class, and model. Here, we propose MetaCAM, an ensemble-based method for combining multiple existing CAM methods based on the consensus of the top-k% most highly activated pixels across component CAMs. We perform experiments to quantifiably determine the optimal combination of 11 CAMs for a given MetaCAM experiment. A new method denoted Cumulative Residual Effect (CRE) is proposed to summarize large-scale ensemble-based experiments. We also present adaptive thresholding and demonstrate how it can be applied to individual CAMs to improve their performance, measured using pixel perturbation method Remove and Debias (ROAD). Lastly, we show that MetaCAM outperforms existing CAMs and refines the most salient regions of images used for model predictions. In a specific example, MetaCAM improved ROAD performance to 0.393 compared to 11 individual CAMs with ranges from -0.101-0.172, demonstrating the importance of combining CAMs through an ensembling method and adaptive thresholding.