AI Paper Atlas

Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning

Few-Shot Learning With Global Class Representations

Aoxue Li, Tiange Luo, Tao Xiang, Weiran Huang, Liwei Wang

ICCV 2019 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Alignment-based few-shot metric learning · 120 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add registration module and sample synthesis together (ablation)global class representation baseline B · miniImageNet 5-way (ablation)accuracyImproved
add sample synthesis alone (ablation)global representation baseline B · miniImageNet 5-way (ablation)accuracyNo clear effect
add registration module alone (ablation)global representation baseline B · miniImageNet 5-way (ablation)accuracyNo clear effect
use soft weighted selection instead of argmax in registration (ablation)registration module · miniImageNetaccuracy: 100% registration accuracyNo clear effect
learn global class representations with joint base-novel training instead of episodic prototypesPrototypical Nets · miniImageNet 5-way 5-shotaccuracy: 68.20 -> 72.34Improved
global representation approach vs strongest baseline on 1-shotMMN · miniImageNet 5-way 1-shotaccuracy: 53.37 vs 53.21No clear effect
global representation approach vs MGMG · Omniglot 20-way 1-shotaccuracy: 97.64 -> 99.63Improved
global representation approach vs MGMG · Omniglot 5-way 5-shotaccuracy: 99.86 -> 99.90No clear effect
use global class representations instead of episodic onesPN · miniImageNet generalizednovel class accuracy: 0.52 -> 12.98Improved
use global class representations instead of episodic onesRN · miniImageNet generalizedoverall accuracy: 32.48 -> 39.14Improved
fix 100 seen class representations and add 20 new novel classesOurs vs PN · miniImageNet + 20 new classesnovel class accuracy: 1.07 -> 12.60Improved
global representation approach vs MNMN · miniImageNet generalizednovel class accuracy: 0.75 -> 12.98Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy