Universal Representation Learning from Multiple Domains for Few-shot Classification

Wei-Hong Li, Xialei Liu, Hakan Bilen
3/25/2021
cs.CV

Official code: VICO-UoE/URL (147 stars, MIT, last commit 2024-02-23). 1 more implementation below.

Abstract

In this paper, we look at the problem of few-shot classification that aims to learn a classifier for previously unseen classes and domains from few labeled samples. Recent methods use adaptation networks for aligning their features to new domains or select the relevant features from multiple domain-specific feature extractors. In this work, we propose to learn a single set of universal deep representations by distilling knowledge of multiple separately trained networks after co-aligning their features with the help of adapters and centered kernel alignment. We show that the universal representations can be further refined for previously unseen domains by an efficient adaptation step in a similar spirit to distance learning methods. We rigorously evaluate our model in the recent Meta-Dataset benchmark and demonstrate that it significantly outperforms the previous methods while being more efficient. Our code will be available at https://github.com/VICO-UoE/URL.

View on arXivView PDF

Code Implementations(2)

VICO-UoE/URLOfficial100%
MIT

Universal Representation Learning from Multiple Domains for Few-shot Classification - ICCV 2021, Cross-domain Few-shot Learning with Task-specific Adapters - CVPR 2022

14717Shell, PythonMar 25, 20212 years agoMIT
few-shot-learningknowledge-distillationmulti-domain-learningmulti-domain-meta-training
google-research/meta-datasetCommunity100%
TensorFlowApache-2.0

A dataset of datasets for learning to learn from few examples

803140Python, Jupyter NotebookMar 5, 20192 years agoApache-2.0
benchmarkfew-shot-learningmachine-learningmamlmeta-learning+1 more

Similar Papers

A Closer Look at Few-shot Classification

88% match1 repo1.2k2019

Cross-domain Few-shot Learning with Task-specific Adapters

88% match2 repos8032021

Prototypical Networks for Few-shot Learning

87% match1 repo1.2k2017

What Makes for Effective Few-shot Point Cloud Classification?

87% match1 repo172023

High-level semantic feature matters few-shot unsupervised domain adaptation

86% match1 repo32023

Cross-Level Distillation and Feature Denoising for Cross-Domain Few-Shot Classification

86% match1 repo282023

Cite this paper

@article{li2021universal,
  title  = {Universal Representation Learning from Multiple Domains for Few-shot Classification},
  author = {Wei-Hong Li and Xialei Liu and Hakan Bilen},
  year   = {2021},
  eprint = {2103.13841},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/2103.13841}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2103.13841.svg)](https://codeofpaper.com/paper/2103.13841)

Discussion