List your company
ResearchFormerly WizWand

50-shot 5-way Classification on miniImageNet (test)

75.54AccuracyMAML
24.923238.064151.20564.3459Apr 13, 2020
Updated 8mo ago

Evaluation results

MethodLinks
MAMLClustering=N/A, Training Protocol=Supervised (Upper Bound), Backbone=4-layer CNN2020.0475.54
ProtoNetsClustering=N/A, Training Protocol=Supervised (Upper Bound), Backbone=4-layer CNN2020.0472.04
CACTUs-MAMLClustering=DeepCluster, Training Protocol=Unsupervised, Backbone=4-layer CNN2020.0469.64
ULDA-MetaOptNetClustering=N/A, Training Protocol=Unsupervised, Backbone=4-layer CNN2020.0467.65
UMTRAClustering=N/A, Training Protocol=Unsupervised, Backbone=4-layer CNN2020.0467.15
ULDA-ProtoNetsClustering=N/A, Training Protocol=Unsupervised, Backbone=4-layer CNN2020.0466.43
linear classifierClustering=DeepCluster, Training Protocol=Unsupervised, Backbone=4-layer CNN2020.0465.28
knn-nearest neighborsClustering=DeepCluster, Training Protocol=Unsupervised, Backbone=4-layer CNN2020.0463.9
CACTUs-ProtoNetsClustering=DeepCluster, Training Protocol=Unsupervised, Backbone=4-layer CNN2020.0463.55
MLP with dropoutClustering=DeepCluster, Training Protocol=Unsupervised, Backbone=4-layer CNN2020.0460.95
Training from scratchClustering=N/A, Training Protocol=Unsupervised, Backbone=4-layer CNN2020.0459.63
cluster matchingClustering=DeepCluster, Training Protocol=Unsupervised, Backbone=4-layer CNN2020.0426.87