Rectifying the Shortcut Learning of Background for Few-Shot Learning

About

The category gap between training and evaluation has been characterised as one of the main obstacles to the success of Few-Shot Learning (FSL). In this paper, we for the first time empirically identify image background, common in realistic images, as a shortcut knowledge helpful for in-class classification but ungeneralizable beyond training categories in FSL. A novel framework, COSOC, is designed to tackle this problem by extracting foreground objects in images at both training and evaluation without any extra supervision. Extensive experiments carried on inductive FSL tasks demonstrate the effectiveness of our approaches.

Xu Luo, Longhui Wei, Liangjian Wen, Jinrong Yang, Lingxi Xie, Zenglin Xu, Qi Tian · 2021

Related benchmarks

Task	Dataset	Result
Few-shot classification	tieredImageNet (test)	Accuracy—	282
5-way Few-shot Classification	MiniImagenet	Accuracy (5-shot)85.16	150
5-way Few-shot Classification	Mini-Imagenet (test)	1-shot Accuracy69.28	141
5-way 5-shot Classification	miniImageNet (test)	Accuracy85.16	56
5-way Few-shot Classification	tieredImageNet	Accuracy (1-shot)73.57	49
5-way 1-shot Classification	Mini-Imagenet (test)	Accuracy69.28	43
5-way Few-shot Classification	tiered-ImageNet (test)	1-shot Acc73.57	33
Image Classification	tiered-ImageNet	1-Shot Acc73.57	25
Image Classification	MiniImagenet	1-Shot Accuracy69.28	16