A Baseline for Few-Shot Image Classification

TL;DR

This paper demonstrates that simple fine-tuning of a deep network can serve as a strong baseline for few-shot image classification, outperforming state-of-the-art methods on multiple datasets and highlighting limitations of current benchmarks.

Contribution

It introduces a straightforward fine-tuning approach as a strong baseline and proposes a metric to quantify the difficulty of few-shot episodes.

Findings

Fine-tuning outperforms current state-of-the-art on standard datasets.

Using many meta-training classes yields high accuracy even with many few-shot classes.

Proposes a new metric to measure the hardness of few-shot episodes.

Abstract

Fine-tuning a deep network trained with the standard cross-entropy loss is a strong baseline for few-shot learning. When fine-tuned transductively, this outperforms the current state-of-the-art on standard datasets such as Mini-ImageNet, Tiered-ImageNet, CIFAR-FS and FC-100 with the same hyper-parameters. The simplicity of this approach enables us to demonstrate the first few-shot learning results on the ImageNet-21k dataset. We find that using a large number of meta-training classes results in high few-shot accuracies even for a large number of few-shot classes. We do not advocate our approach as the solution for few-shot learning, but simply use the results to highlight limitations of current benchmarks and few-shot protocols. We perform extensive studies on benchmark datasets to propose a metric that quantifies the "hardness" of a few-shot episode. This metric can be used to report the performance of few-shot algorithms in a more systematic way.