Rethinking Few-Shot Image Classification: a Good Embedding Is All You Need?

TL;DR

This paper demonstrates that a simple approach of learning a strong embedding and training a linear classifier outperforms complex meta-learning algorithms in few-shot image classification, suggesting a need to rethink current benchmarks.

Contribution

The study shows that a straightforward embedding-based method surpasses state-of-the-art meta-learning techniques in few-shot classification, challenging existing assumptions.

Findings

Simple embedding + linear classifier outperforms meta-learning methods

Self-distillation further improves performance

Motivates re-evaluation of few-shot learning benchmarks

Abstract

The focus of recent meta-learning research has been on the development of learning algorithms that can quickly adapt to test time tasks with limited data and low computational cost. Few-shot learning is widely used as one of the standard benchmarks in meta-learning. In this work, we show that a simple baseline: learning a supervised or self-supervised representation on the meta-training set, followed by training a linear classifier on top of this representation, outperforms state-of-the-art few-shot learning methods. An additional boost can be achieved through the use of self-distillation. This demonstrates that using a good learned embedding model can be more effective than sophisticated meta-learning algorithms. We believe that our findings motivate a rethinking of few-shot image classification benchmarks and the associated role of meta-learning algorithms. Code is available at: http://github.com/WangYueFt/rfs/.