List your company
ResearchFormerly WizWand

5-Shot 5-Way Classification on MiniImageNet (test)

73.94AccuracyTrainProto-FSL
36.39646.14355.8965.637Dec 10, 2018Jul 1, 2019Jan 20, 2020Aug 10, 2020Mar 1, 2021Sep 20, 2021Apr 12, 2022
Updated 8mo ago

Evaluation results

MethodLinks
TrainProto-FSLSupervision=unsupervised2021.0673.94
FEAT‡Setup=Transductive, Backbone=ConvNet, Prediction Strategy=via Semi-ProtoNet2018.1272.89
FEAT†Setup=Transductive, Backbone=ConvNet, Prediction Strategy=via ProtoNet2018.1272.65
TEAMSetup=Transductive, Backbone=ConvNet2018.1272.04
Semi-ProtoNet (Ours)Setup=Transductive, Backbone=ConvNet2018.1271.76
FEATSetup=Standard, Backbone=ConvNet2018.1271.61
ProtoNetSetup=Standard, Backbone=ConvNet2018.1271.33
TPNSetup=Transductive, Backbone=ConvNet2018.1269.86
CSSL-FSL_Image168Supervision=unsupervised2021.0668.91
ProtoNetsSupervision=supervised2021.0667.9
Semi-ProtoNetSetup=Transductive, Backbone=ConvNet2018.1264.39
OracleNoise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot2022.0462.6
MAMLSupervision=supervised2021.0660
ULDA-ProtoNetsSupervision=unsupervised2021.0655.41
ULDA-MetaOptNetSupervision=unsupervised2021.0654.49
CACTUS-MAMLSupervision=unsupervised2021.0653.97
TraNFS-3Noise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot, Layers=32022.0453.96
CACTUS-ProtoNetsSupervision=unsupervised2021.0653.36
UMTRASupervision=unsupervised2021.0650.73
TraNFS-2Noise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot, Layers=22022.0450.63
Similarity weighted prototypes (Cosine)Noise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot, distance=cosine2022.0449.4
Similarity weighted prototypes (Absolute)Noise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot, distance=absolute2022.0449.38
AAL-MAML++Supervision=unsupervised2021.0649.18
Spatial median prototypes (Median)Noise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot2022.0448.81
Similarity weighted prototypes (Euclidean)Noise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot, distance=euclidean2022.0448.67
RNNPNoise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot2022.0447.88
Baseline++Noise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot2022.0447.82
Vanilla ProtoNetNoise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot2022.0447.77
UFLSTSupervision=unsupervised2021.0645.03
Linear ClassifierNoise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot2022.0444.49
Matching NetworksNoise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot2022.0443.53
MAMLNoise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot2022.0440.67
Nearest k = 5Noise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot2022.0440.39
AAL-ProtoNetsSupervision=unsupervised2021.0640.29
Nearest k = 1Noise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot2022.0437.97
Nearest k = 3Noise Proportion=40%, Noise Type=paired label swap, Shot-Way Configuration=5-way 5-shot2022.0437.84