LFI-CAM: Learning Feature Importance for Better Visual Explanation

TL;DR

LFI-CAM introduces an end-to-end trainable model that simultaneously improves CNN classification accuracy and generates more reliable, stable visual explanations by learning feature importance rather than direct attention maps.

Contribution

The paper proposes LFI-CAM, a novel architecture that learns feature importance for better visual explanations and classification performance in CNNs, addressing limitations of previous CAM-based methods.

Findings

LFI-CAM outperforms baseline models in classification accuracy.

LFI-CAM produces higher quality and more stable attention maps.

The model enhances feature representation focusing on important image features.

Abstract

Class Activation Mapping (CAM) is a powerful technique used to understand the decision making of Convolutional Neural Network (CNN) in computer vision. Recently, there have been attempts not only to generate better visual explanations, but also to improve classification performance using visual explanations. However, the previous works still have their own drawbacks. In this paper, we propose a novel architecture, LFI-CAM, which is trainable for image classification and visual explanation in an end-to-end manner. LFI-CAM generates an attention map for visual explanation during forward propagation, at the same time, leverages the attention map to improve the classification performance through the attention mechanism. Our Feature Importance Network (FIN) focuses on learning the feature importance instead of directly learning the attention map to obtain a more reliable and consistent attention map. We confirmed that LFI-CAM model is optimized not only by learning the feature importance but also by enhancing the backbone feature representation to focus more on important features of the input image. Experimental results show that LFI-CAM outperforms the baseline models's accuracy on the classification tasks as well as significantly improves on the previous works in terms of attention map quality and stability over different hyper-parameters.