List your company
ResearchFormerly WizWand

Few-shot classification on CUB (test) (Meta-trained on Mini-ImageNet)

80.92Accuracy (5-shot)FCAM
46.475255.417664.3673.3024Jun 6, 2020Feb 8, 2021Oct 14, 2021Jun 19, 2022Feb 22, 2023Oct 28, 2023Jul 2, 2024
Updated 7mo ago

Evaluation results

MethodLinks
FCAM2024.0780.92—58.2—
TRIDENTBackbone=Conv4, Approach=transductive feature extraction2022.0880.74—84.61—
PEME-BMSBackbone=WRN, Approach=transductive feature extraction and inference2022.0879.15—63.9—
DeepEMD v22024.0778.86—54.24—
MCL-KatzBackbone=ResNet-122021.0677.39—53.22—
FRNBackbone=RN12, Approach=inductive inference2022.0877.09—54.11—
PT-MAPSetting=Transductive, Backbone=WRN2020.0676.51—62.49—
PT+MAPBackbone=WRN, Approach=transductive feature extraction and inference2022.0876.51—62.49—
MCLBackbone=ResNet-122021.0675.93—55.48—
Transfer+SGCSetting=Transductive, Backbone=WRN2020.0673.46—58.63—
linear classifier2024.0773.3—50.37—
FRNBackbone=ResNet-122021.0672.97—51.6—
Assoc-AlignBackbone=WRN, Approach=transductive feature extraction2022.0872.37—47.25—
ProtoNet2024.0772.02—50.01—
MetaQDABackbone=WRN, Approach=transductive feature extraction2022.0871.84—53.75—
MetaQDABackbone=WRN2021.0171.84—53.75—
KNN2024.0771.25—50.84—
FEATBackbone=RN18, Approach=transductive feature extraction2022.0871.08—50.67—
TIM-GDBackbone=WRN, Approach=transductive inference2022.0871———
S2M2_RSetting=Inductive, Backbone=WRN2020.0670.44—48.24—
S2M2Backbone=WRN, Approach=transductive feature extraction2022.0870.44—48.24—
S2M2Backbone=WRN2021.0170.44—48.24—
Centroid et al.Backbone=ResNet-182021.0670.37—46.85—
PT-NCMSetting=Inductive, Backbone=WRN2020.0670.22—48.37—
Neg-marginBackbone=ResNet-182021.0669.3———
MatchNet2024.0769.14—51.65—
cosine classifier2024.0769.01—44.17—
MetaQDABackbone=ResNet-182021.0168.59—48.88—
GNN+FTBackbone=ResNet-102021.0666.98—47.47—
LFWTBackbone=RN10, Approach=transductive feature extraction and inference2022.0866.98—47.47—
LFWTBackbone=ResNet-102021.0166.98—47.47—
SIMPLESHOT†Backbone=WRN2021.0166.77—49.65—
LRP (CAN)Backbone=ResNet-122021.0166.58—46.23—
LaplacianShotBackbone=WRN, Approach=transductive inference2022.0866.33—55.46—
Manifold MixupSetting=Inductive, Backbone=WRN2020.0666.03—46.21—
SimpleShotBackbone=WRN, Approach=inductive inference2022.0865.63—48.56—
BaselineBackbone=ResNet-102021.0665.57———
SIMPLESHOT†Backbone=ResNet-182021.0165.56—46.68—
MetaOptNetBackbone=ResNet-122021.0664.98—44.79—
ProtoNet+MCLBackbone=ResNet-122021.0664.76—42.02—
LRP (GNN)Backbone=ResNet-102021.0164.44—48.29—
MetaQDABackbone=Conv-42021.0164.4—47.25—
BaselineBackbone=WRN-28-10, Training Paradigm=Supervised, Evaluation Protocol=5-way2022.1062.12—42.89—
Baseline++Backbone=ResNet-182021.0662.04———
Baseline++Backbone=RN18, Approach=transductive feature extraction2022.0862.04—42.85—
BASELINE++Backbone=ResNet-182021.0162.04—42.85—
PROTONETBackbone=ResNet-182021.0162.02———
SIMPLESHOT†Backbone=Conv-42021.0161.44—45.36—
SIBBackbone=WRN-28-10, Training Paradigm=Supervised, Evaluation Protocol=5-way2022.1059.94—43.27—
BECLR2024.0259.51—43.45—
RelationNet+FTBackbone=ResNet-102021.0659.46—44.07—
Deep Laplacian eigenmapsBackbone=WRN-28-10, Training Paradigm=Unsupervised, Evaluation Protocol=5-way2022.1058.86—41.08—
Laplacian Eigenmaps2024.0258.86—41.08—
HMS2024.0258.32—40.75—
RELATIONNETBackbone=ResNet-182021.0157.71———
PsCo2024.0257.38———
OVE(PL)Backbone=Conv4, Approach=inductive inference2022.0857.23—37.49—
Barlow TwinsBackbone=WRN-28-10, Training Paradigm=Unsupervised, Evaluation Protocol=5-way2022.1057.16—40.46—
Barlow Twins2024.0257.16—40.46—
BYOLBackbone=WRN-28-10, Training Paradigm=Unsupervised, Evaluation Protocol=5-way2022.1056.92—40.63—
BYOL2024.0256.92—40.63—
MTLBackbone=WRN-28-10, Training Paradigm=Supervised, Evaluation Protocol=5-way2022.1056.89—43.15—
LEOBackbone=WRN-28-10, Training Paradigm=Supervised, Evaluation Protocol=5-way2022.1056.66—41.45—
Baseline++Setting=Inductive, Backbone=WRN2020.0656.64—40.44—
MoCo v2Backbone=WRN-28-10, Training Paradigm=Unsupervised, Evaluation Protocol=5-way2022.1056.49—39.29—
MoCo v22024.0256.49—39.29—
SimCLRBackbone=WRN-28-10, Training Paradigm=Unsupervised, Evaluation Protocol=5-way2022.1055.89—38.25—
SimCLR2024.0255.89—38.25—
DKT+CosBackbone=Conv4, Approach=inductive inference2022.0855.65—40.22—
Meta-GMVAEBackbone=WRN-28-10, Training Paradigm=Unsupervised, Evaluation Protocol=5-way2022.1055.65—38.09—
Meta-GMVAE2024.0255.65—38.09—
ProtoNetBackbone=ResNet-122021.0655.29—40.05—
MatchingNet+FTBackbone=ResNet-102021.0655.23—36.61—
MAMLBackbone=WRN-28-10, Training Paradigm=Supervised, Evaluation Protocol=5-way2022.1055.04—39.06—
NNCLRreproduced=true2024.0254.78—39.37—
SwAVreproduced=true2024.0253.94—38.34—
Matching NetBackbone=WRN-28-10, Training Paradigm=Supervised, Evaluation Protocol=5-way2022.1053.08—42.04—
MAMLBackbone=ResNet-182021.0151.34———
UMTRABackbone=WRN-28-10, Training Paradigm=Unsupervised, Evaluation Protocol=5-way2022.1050.21—33.59—
CACTUS-MAMLBackbone=WRN-28-10, Training Paradigm=Unsupervised, Evaluation Protocol=5-way2022.1049.97—33.48—
MAMLBackbone=Conv4, Approach=inductive inference2022.0848.83—34.01—
ABMLBackbone=Conv4, Approach=inductive inference2022.0847.8—31.51—
ANILShots=1-shot, Ways=5-way, Backbone=ResNet-122022.10—37.3——
ANILShots=5-shot, Ways=5-way, Backbone=ResNet-122022.10—53.42——
ANIL + SiMTShots=1-shot, Ways=5-way, Backbone=ResNet-122022.10—38.86——
ANIL + SiMTShots=5-shot, Ways=5-way, Backbone=ResNet-122022.10—56.03——
Baselineshots=5-shot, backbone=ResNet-182020.03—65.57——
BaselineBackbone=ResNet-18, Shot=5-shot2020.08—65.6——
BaselineBackbone=ResNet-18, Shots=5-shot2023.04—65.57——
Baseline++shots=5-shot, backbone=ResNet-182020.03—62.04——
Baseline++Shots=1-shot, Ways=5-ways, Backbone=Conv-4, Number of runs=32019.10—39.19——
Baseline++Backbone=ResNet-18, Shots=5-shot2023.04—62.04——
BASELINE++Backbone=Conv-42021.01——39.19—
DKTKernel=Linear, Shots=1-shot, Ways=5-ways, Backbone=Conv-4, Number of runs=32019.10—38.72——
DKTKernel=CosSim, Shots=1-shot, Ways=5-ways, Backbone=Conv-4, Number of runs=32019.10—40.22——
DKTKernel=BNCosSim, Shots=1-shot, Ways=5-ways, Backbone=Conv-4, Number of runs=32019.10—40.14——
DKTBackbone=Conv-42021.01——40.22—
ESPTBackbone=ResNet-12, Shots=1-shot, Local image representation=true2023.04—54.14——
ESPTBackbone=ResNet-12, Shots=5-shot, Local image representation=true2023.04—74.91——
FEATBackbone=ResNet-12, Shots=1-shot2023.04—50.67——