TAME: Attention Mechanism Based Feature Fusion for Generating Explanation Maps of Convolutional Neural Networks

TL;DR

TAME introduces a trainable attention-based method for generating explanation maps of CNNs, improving interpretability by effectively combining multi-layer features into clear visual explanations.

Contribution

The paper proposes TAME, a novel attention mechanism that efficiently produces explanation maps for any CNN, streamlining the process and enhancing interpretability.

Findings

TAME outperforms previous explanation methods on VGG-16 and ResNet-50.

Explanation maps can be generated in a single forward pass after training.

Comprehensive ablation study validates the effectiveness of TAME's architecture.

Abstract

The apparent ``black box'' nature of neural networks is a barrier to adoption in applications where explainability is essential. This paper presents TAME (Trainable Attention Mechanism for Explanations), a method for generating explanation maps with a multi-branch hierarchical attention mechanism. TAME combines a target model's feature maps from multiple layers using an attention mechanism, transforming them into an explanation map. TAME can easily be applied to any convolutional neural network (CNN) by streamlining the optimization of the attention mechanism's training method and the selection of target model's feature maps. After training, explanation maps can be computed in a single forward pass. We apply TAME to two widely used models, i.e. VGG-16 and ResNet-50, trained on ImageNet and show improvements over previous top-performing methods. We also provide a comprehensive ablation study comparing the performance of different variations of TAME's architecture. TAME source code is made publicly available at https://github.com/bmezaris/TAME