Time for a Background Check! Uncovering the impact of Background Features on Deep Neural Networks
This addresses the problem of understanding feature reliance in DNNs for image classification, which is incremental as it builds on existing studies of network behavior.
The paper investigated how background features impact deep neural networks' performance, finding that more expressive networks rely more on background features but also maintain accuracy when backgrounds are altered.
With increasing expressive power, deep neural networks have significantly improved the state-of-the-art on image classification datasets, such as ImageNet. In this paper, we investigate to what extent the increasing performance of deep neural networks is impacted by background features? In particular, we focus on background invariance, i.e., accuracy unaffected by switching background features and background influence, i.e., predictive power of background features itself when foreground is masked. We perform experiments with 32 different neural networks ranging from small-size networks to large-scale networks trained with up to one Billion images. Our investigations reveal that increasing expressive power of DNNs leads to higher influence of background features, while simultaneously, increases their ability to make the correct prediction when background features are removed or replaced with a randomly selected texture-based background.