Rethinking the Image Feature Biases Exhibited by Deep CNN Models

TL;DR

This paper investigates feature biases in CNNs, revealing that models can be directed to focus on specific features depending on task design, and that combined features often have greater influence than individual ones.

Contribution

It introduces a method to identify and manipulate feature biases in CNNs by designing tasks that emphasize particular features, advancing understanding of feature importance in deep models.

Findings

Combined features exert more influence than single features.

Neural models can be biased toward specific features based on task design.

Different tasks lead to different feature biases in CNNs.

Abstract

In recent years, convolutional neural networks (CNNs) have been applied successfully in many fields. However, such deep neural models are still regarded as black box in most tasks. One of the fundamental issues underlying this problem is understanding which features are most influential in image recognition tasks and how they are processed by CNNs. It is widely accepted that CNN models combine low-level features to form complex shapes until the object can be readily classified, however, several recent studies have argued that texture features are more important than other features. In this paper, we assume that the importance of certain features varies depending on specific tasks, i.e., specific tasks exhibit a feature bias. We designed two classification tasks based on human intuition to train deep neural models to identify anticipated biases. We devised experiments comprising many tasks to test these biases for the ResNet and DenseNet models. From the results, we conclude that (1) the combined effect of certain features is typically far more influential than any single feature; (2) in different tasks, neural models can perform different biases, that is, we can design a specific task to make a neural model biased toward a specific anticipated feature.