A Universal Representation Transformer Layer for Few-Shot Image Classification

TL;DR

This paper introduces a Universal Representation Transformer layer that meta-learns to effectively combine diverse domain-specific features, achieving state-of-the-art results in multi-domain few-shot image classification.

Contribution

The paper proposes the URT layer, a novel meta-learning module that dynamically re-weights and composes features from multiple domains for improved few-shot classification.

Findings

URT achieves state-of-the-art on Meta-Dataset.

URT performs well across many data sources.

Visualization shows effective cross-domain generalization.

Abstract

Few-shot classification aims to recognize unseen classes when presented with only a small number of samples. We consider the problem of multi-domain few-shot image classification, where unseen classes and examples come from diverse data sources. This problem has seen growing interest and has inspired the development of benchmarks such as Meta-Dataset. A key challenge in this multi-domain setting is to effectively integrate the feature representations from the diverse set of training domains. Here, we propose a Universal Representation Transformer (URT) layer, that meta-learns to leverage universal features for few-shot classification by dynamically re-weighting and composing the most appropriate domain-specific representations. In experiments, we show that URT sets a new state-of-the-art result on Meta-Dataset. Specifically, it achieves top-performance on the highest number of data sources compared to competing methods. We analyze variants of URT and present a visualization of the attention score heatmaps that sheds light on how the model performs cross-domain generalization. Our code is available at https://github.com/liulu112601/URT.