List your company
ResearchFormerly WizWand

Few-shot Image Classification on miniImageNet meta (test)

93.21AccuracyNAO-centroid
41.449254.887168.32581.7629Apr 7, 2019Dec 17, 2019Aug 28, 2020May 10, 2021Jan 20, 2022Oct 2, 2022Jun 14, 2023
Updated 8mo ago

Evaluation results

MethodLinks
NAO-centroidFew-shot setting=5-way 5-shot2023.0693.21——
SeedSelectFew-shot setting=5-way 5-shot2023.0692.08——
DiffAlignFew-shot setting=5-way 5-shot, pre-trained on external datasets=true2023.0688.63——
FeLMiFew-shot setting=5-way 5-shot2023.0686.08——
Textual InversionFew-shot setting=5-way 5-shot, results from=[51], pre-trained on external datasets=true2023.0685.44——
Stable DiffusionFew-shot setting=5-way 5-shot, pre-trained on external datasets=true2023.0685.05——
SVAEFew-shot setting=5-way 5-shot, pre-trained on external datasets=true2023.0680.7——
MetaOptNet-SVM-trainvalbackbone=ResNet-12, training_set=meta-training + meta-validation2019.0480——
SEGAFew-shot setting=5-way 5-shot, multi-modal (class labels)=true2023.0679.03——
MetaOptNet-SVMbackbone=ResNet-122019.0478.63——
MetaOptNet-RRbackbone=ResNet-122019.0477.88——
LEObackbone=WRN-28-102019.0477.59——
TADAMbackbone=ResNet-122019.0476.7——
Activation to Parameterbackbone=WRN-28-102019.0473.74——
Dynamic Few-shotbackbone=64-64-128-1282019.0473——
AdaResNetbackbone=ResNet-122019.0471.94——
Transductive Prop Netsbackbone=64-64-64-642019.0469.86——
SNAILbackbone=ResNet-122019.0468.88——
R2D2backbone=96-192-384-5122019.0468.8——
Prototypical Networksbackbone=64-64-64-642019.0468.2——
Label-HallucFew-shot setting=5-way 5-shot2023.0667.04——
Relation Networksbackbone=64-96-128-2562019.0465.32——
MetaOptNet-SVM-trainvalbackbone=ResNet-12, training_set=meta-training + meta-validation2019.0464.09——
MAMLbackbone=32-32-32-322019.0463.11——
MetaOptNet-SVMbackbone=ResNet-122019.0462.64——
LEObackbone=WRN-28-102019.0461.76——
MetaOptNet-RRbackbone=ResNet-122019.0461.41——
Meta-Learning LSTMbackbone=64-64-64-642019.0460.6——
Activation to Parameterbackbone=WRN-28-102019.0459.6——
TADAMbackbone=ResNet-122019.0458.5——
AdaResNetbackbone=ResNet-122019.0456.88——
Dynamic Few-shotbackbone=64-64-128-1282019.0456.2——
SNAILbackbone=ResNet-122019.0455.71——
Transductive Prop Netsbackbone=64-64-64-642019.0455.51——
Matching Networksbackbone=64-64-64-642019.0455.31——
R2D2backbone=96-192-384-5122019.0451.2——
Relation Networksbackbone=64-96-128-2562019.0450.44——
ReptileN-way=5, K-shot=12019.0949.97——
Prototypical Networksbackbone=64-64-64-642019.0449.42——
iMAML HFN-way=5, K-shot=1, Inner loop optimizer=Hessian-free, CG steps=5, lambda=0.52019.0949.3——
iMAML GDN-way=5, K-shot=1, Inner loop gradient steps=10, CG steps=5, lambda=0.52019.0948.96——
MAMLN-way=5, K-shot=12019.0948.7——
MAMLbackbone=32-32-32-322019.0448.7——
first-order MAMLN-way=5, K-shot=12019.0948.07——
Matching Networksbackbone=64-64-64-642019.0443.56——
Meta-Learning LSTMbackbone=64-64-64-642019.0443.44——
Fine-tuningBackbone=ResNet-12, Number of meta-test tasks=6002021.11—60.5676.42
l2-NormBackbone=ResNet-12, Number of meta-test tasks=6002021.11—60.9376.57
l2-PGMBackbone=ResNet-12, Number of meta-test tasks=6002021.11—61.3577.33
l2-SPBackbone=ResNet-12, Number of meta-test tasks=6002021.11—61.4877.02
LSBackbone=ResNet-12, Number of meta-test tasks=6002021.11—61.3176.73
REGSLBackbone=ResNet-12, Number of meta-test tasks=6002021.11—61.7178.03