Texture Bias Of CNNs Limits Few-Shot Classification Performance

TL;DR

This paper investigates how the texture bias of CNNs negatively impacts few-shot image classification and introduces a bias correction method that achieves state-of-the-art results on miniImageNet.

Contribution

It identifies the detrimental effect of texture bias on few-shot learning and proposes a simple correction method that improves performance significantly.

Findings

Texture bias harms few-shot classification accuracy.

Correcting texture bias leads to state-of-the-art results on miniImageNet.

The proposed method is simpler than existing approaches.

Abstract

Accurate image classification given small amounts of labelled data (few-shot classification) remains an open problem in computer vision. In this work we examine how the known texture bias of Convolutional Neural Networks (CNNs) affects few-shot classification performance. Although texture bias can help in standard image classification, in this work we show it significantly harms few-shot classification performance. After correcting this bias we demonstrate state-of-the-art performance on the competitive miniImageNet task using a method far simpler than the current best performing few-shot learning approaches.