TL;DR: Transductive fine-tuning of a deep network is a strong baseline for few-shot image classification and outperforms the state-of-the-art on all standard benchmarks.
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.
Keywords: few-shot learning, transductive learning, fine-tuning, baseline, meta-learning
Data: [CIFAR-FS](https://paperswithcode.com/dataset/cifar-fs), [FC100](https://paperswithcode.com/dataset/fc100), [ImageNet](https://paperswithcode.com/dataset/imagenet), [mini-Imagenet](https://paperswithcode.com/dataset/mini-imagenet), [tieredImageNet](https://paperswithcode.com/dataset/tieredimagenet)
Code: [![github](/images/github_icon.svg) amazon-science/few-shot-baseline](https://github.com/amazon-science/few-shot-baseline) + [![Papers with Code](/images/pwc_icon.svg) 2 community implementations](https://paperswithcode.com/paper/?openreview=rylXBkrYDS)
Community Implementations: [![CatalyzeX](/images/catalyzex_icon.svg) 1 code implementation](https://www.catalyzex.com/paper/arxiv:1909.02729/code)
Original Pdf: pdf