Object-Centric Cropping for Visual Few-Shot Classification

TL;DR

This paper introduces an object-centric cropping method that leverages minimal local information, such as a pixel, to improve few-shot image classification performance, utilizing models like Segment Anything or unsupervised extraction.

Contribution

It presents a novel approach that enhances few-shot classification by incorporating local object positioning, using minimal supervision with advanced segmentation models.

Findings

Significant accuracy improvements on benchmark datasets.

Object-centric cropping reduces background noise in classification.

Minimal input, like a single pixel, suffices for effective segmentation.

Abstract

In the domain of Few-Shot Image Classification, operating with as little as one example per class, the presence of image ambiguities stemming from multiple objects or complex backgrounds can significantly deteriorate performance. Our research demonstrates that incorporating additional information about the local positioning of an object within its image markedly enhances classification across established benchmarks. More importantly, we show that a significant fraction of the improvement can be achieved through the use of the Segment Anything Model, requiring only a pixel of the object of interest to be pointed out, or by employing fully unsupervised foreground object extraction methods.