Home Knowledge Base Meta-Dataset

Meta-Dataset is a large-scale benchmark for evaluating few-shot learning algorithms, consisting of a diverse collection of datasets spanning different visual domains. Introduced by Triantafillou et al. (2020), it addressed critical limitations of earlier single-domain evaluations.

Why Meta-Dataset Was Needed

Component Datasets (10 Domains)

DomainDatasetClassesDescription
Natural ImagesImageNet1,000General object recognition
HandwritingOmniglot1,623Handwritten characters from 50 alphabets
AircraftFGVC-Aircraft100Fine-grained aircraft model recognition
BirdsCUB-200200Fine-grained bird species
TexturesDTD47Describable texture patterns
DrawingsQuick Draw345Hand-drawn sketches
FungiFGVCx Fungi1,394Mushroom species identification
FlowersVGG Flower102Flower species recognition
SignsTraffic Signs43Traffic sign classification
ObjectsMSCOCO80Object categories in context

Key Design Innovations

Evaluation Protocol

Key Findings

Meta-Dataset established the gold standard for few-shot learning evaluation — any new few-shot method must demonstrate effectiveness across its diverse domains to be considered truly general.

meta-datasetfew-shot learning

Explore 500+ Semiconductor & AI Topics

From EUV lithography to CUDA optimization — search the full knowledge base or chat with our AI assistant.