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.
@article{arxiv.1910.08519,
title = {Texture Bias Of CNNs Limits Few-Shot Classification Performance},
author = {Sam Ringer and Will Williams and Tom Ash and Remi Francis and David MacLeod},
journal= {arXiv preprint arXiv:1910.08519},
year = {2019}
}