Targeted Background Removal Creates Interpretable Feature Visualizations

TL;DR

This paper introduces a background removal training technique that enhances the interpretability of feature visualizations in neural networks by forcing models to focus on main objects, resulting in more recognizable features.

Contribution

The study demonstrates that background removal during training significantly improves the interpretability of learned features in neural networks compared to traditional methods.

Findings

Background removal improves feature visualization clarity.

Models trained with background removal focus on main objects.

Enhanced interpretability aids understanding of learned features.

Abstract

Feature visualization is used to visualize learned features for black box machine learning models. Our approach explores an altered training process to improve interpretability of the visualizations. We argue that by using background removal techniques as a form of robust training, a network is forced to learn more human recognizable features, namely, by focusing on the main object of interest without any distractions from the background. Four different training methods were used to verify this hypothesis. The first used unmodified pictures. The second used a black background. The third utilized Gaussian noise as the background. The fourth approach employed a mix of background removed images and unmodified images. The feature visualization results show that the background removed images reveal a significant improvement over the baseline model. These new results displayed easily recognizable features from their respective classes, unlike the model trained on unmodified data.