Few-Shot Classification on Mini-Imagenet 5-way 5-shot
98AccuracyPMF
Evaluation results
| Method | Links | ||
|---|---|---|---|
| PMFBackbone=ViT-small, Training strategy=Self-Supervised + MT, Extra data=false, Extra labels=false2024.03 | 98 | — | |
| DIPABackbone=ViT-small, Training strategy=Self-Supervised, Extra data=false, Extra labels=false2024.03 | 94.3 | 0.3 | |
| CNAPS + FETIBackbone=WRN-28-10, Training strategy=Transductive, Extra data=false, Extra labels=true2024.03 | 91.5 | 0.4 | |
| PTMAP-SF-SOTptransductive=true, backbone=wrn-28-102022.04 | 91.34 | — | |
| ProtoNetBackbone=AMDIM ResNet, Training strategy=Self-Supervised, Extra data=false, Extra labels=false2024.03 | 91 | 0.1 | |
| PTMAP-SFtransductive=true, backbone=wrn-28-102022.04 | 90.62 | — | |
| PTMAP-SOTttransductive=true, backbone=wrn-28-102022.04 | 90.51 | — | |
| Sill-Net-SOTptransductive=true, backbone=wrn-28-102022.04 | 89.65 | — | |
| PTMAP-SOTptransductive=true, backbone=wrn-28-102022.04 | 89.56 | — | |
| Sill-Nettransductive=true, backbone=wrn-28-102022.04 | 89.14 | — | |
| PTMAPtransductive=true, backbone=wrn-28-102022.04 | 88.8 | — | |
| PT-MAPBackbone=WRN-28-10, Training strategy=Transductive, Extra data=false, Extra labels=true2024.03 | 88.8 | 0.1 | |
| EPNet + SSLBackbone=WRN-28-10, Training strategy=Self-Supervised, Extra data=false, Extra labels=false2024.03 | 88.1 | 0.5 | |
| PTMAP-SOTptransductive=true, backbone=resnet-122022.04 | 86.01 | — | |
| PTMAP-SOTttransductive=true, backbone=resnet-122022.04 | 85.49 | — | |
| PTMAPtransductive=true, backbone=resnet-122022.04 | 85.2 | — | |
| PTMAP-COSINEtransductive=true, backbone=wrn-28-102022.04 | 84.68 | — | |
| FewTureBackbone=ViT-small, Training strategy=Self-Supervised, Extra data=false, Extra labels=false2024.03 | 84.5 | 0.5 | |
| PTMAP-SOFTMAXtransductive=true, backbone=wrn-28-102022.04 | 83.83 | — | |
| PLCMBackbone=ResNet12, Training strategy=Semi-Supervised, Extra data=false, Extra labels=true2024.03 | 83.7 | 0.6 | |
| ProtoNet-SOTttransductive=true, backbone=resnet-122022.04 | 83.09 | — | |
| RENettransductive=false, backbone=resnet-122022.04 | 82.58 | — | |
| DeepEMDtransductive=false, backbone=resnet-122022.04 | 82.41 | — | |
| FEATtransductive=false, backbone=resnet-122022.04 | 82.05 | — | |
| ProtoNet-SOTptransductive=true, backbone=resnet-122022.04 | 81.84 | — | |
| ProtoNettransductive=false, backbone=resnet-122022.04 | 80.33 | — | |
| ProtoNettransductive=false, backbone=wrn-28-102022.04 | 79.97 | — | |
| ProtoNetBackbone=WRN-28-10, Training strategy=Self-Supervised, Extra data=false, Extra labels=false2024.03 | 79.9 | 0.3 | |
| Meta-BaselineBackbone=ResNet12, Training strategy=Inductive, Extra data=false, Extra labels=false2024.03 | 79.3 | 0.2 | |
| SIBBackbone=WRN-28-10, Training strategy=Transductive, Extra data=false, Extra labels=true2024.03 | 79.2 | 0.4 | |
| LSTBackbone=ResNet12, Training strategy=Semi-Supervised, Extra data=false, Extra labels=true2024.03 | 78.7 | 0.8 | |
| MetaOpt-SVMBackbone=ResNet12, Training strategy=Inductive, Extra data=false, Extra labels=false2024.03 | 78.6 | 0.5 | |
| Fine-tuningBackbone=WRN-28-10, Training strategy=Transductive, Extra data=false, Extra labels=true2024.03 | 78.4 | 0.5 | |
| FEATtransductive=false, backbone=conv-42022.04 | 71.61 | — | |
| ProtoNet-SOTttransductive=true, backbone=conv-42022.04 | 70.4 | — | |
| ProtoNet-SOTptransductive=true, backbone=conv-42022.04 | 69.39 | — | |
| RFSChannels=482022.01 | 69.1 | — | |
| MAML++Inner Steps=52018.10 | 68.32 | — | |
| MAML++Channels=482022.01 | 68.3 | — | |
| HTChannels=482022.01 | 68.1 | — | |
| MAML++Training Mode=Supervised Upper Bound2019.02 | 67.87 | — | |
| RFSChannels=322022.01 | 67.1 | — | |
| HTChannels=322022.01 | 67.1 | — | |
| MAML++Channels=322022.01 | 66.8 | — | |
| ProtoNettransductive=false, backbone=conv-42022.04 | 66.79 | — | |
| RelationNettransductive=false, backbone=conv-42022.04 | 66.6 | — | |
| Baseline++Backbone=CNN-4-64, Training strategy=Inductive, Extra data=false, Extra labels=false2024.03 | 66.4 | 0.6 | |
| ProtoNetsTraining Mode=Supervised Upper Bound2019.02 | 65.56 | — | |
| MAML++Channels=162022.01 | 64.6 | — | |
| MAML (local reproduction)Inner Steps=52018.10 | 64.39 | — | |
| HTChannels=162022.01 | 64.2 | — | |
| MAMLTraining Mode=Supervised Upper Bound2019.02 | 64.13 | — | |
| Meta-SGDInner Steps=12018.10 | 64.03 | — | |
| RFSChannels=162022.01 | 63.5 | — | |
| MAML (original paper)Inner Steps=52018.10 | 63.11 | — | |
| MAMLtransductive=false, backbone=conv-42022.04 | 62.71 | — | |
| HTChannels=82022.01 | 59.3 | — | |
| MAML++Channels=82022.01 | 59 | — | |
| RFSChannels=82022.01 | 56.1 | — | |
| Matching Nets2018.10 | 55.31 | — | |
| Units-MLTraining tasks=10000 Tasks2022.03 | 54.23 | — | |
| DeepCluster CACTUS-MAMLLearning Paradigm=Unsupervised, Clusterer=DeepCluster2019.02 | 53.97 | — | |
| DeepCluster CACTUS-ProtoNetsLearning Paradigm=Unsupervised, Clusterer=DeepCluster2019.02 | 53.36 | — | |
| MetaSGD NTSTraining tasks=10000 Tasks2022.03 | 51.69 | — | |
| MetaSGD TSTraining tasks=10000 Tasks2022.03 | 51.54 | — | |
| UMTRA + AutoAugmentLearning Paradigm=Unsupervised, Augmentation=AutoAugment2019.02 | 50.73 | — | |
| MetaSGD TSTraining tasks=8000 Tasks2022.03 | 50.35 | — | |
| Units-NTSTraining tasks=10000 Tasks2022.03 | 49.54 | — | |
| AAL-MAML++Learning Paradigm=Unsupervised, Algorithm=MAML++2019.02 | 49.18 | — | |
| MetaSGD NTSTraining tasks=8000 Tasks2022.03 | 48.95 | — | |
| Units-MLTraining tasks=8000 Tasks2022.03 | 48.34 | — | |
| Units-NTSTraining tasks=8000 Tasks2022.03 | 48.32 | — | |
| Units-MLTraining tasks=4000 Tasks2022.03 | 47.33 | — | |
| Versa TSTraining tasks=10000 Tasks2022.03 | 46.21 | — | |
| MetaSGD TSTraining tasks=4000 Tasks2022.03 | 45.51 | — | |
| Versa NTSTraining tasks=10000 Tasks2022.03 | 45.48 | — | |
| Versa TSTraining tasks=8000 Tasks2022.03 | 45.48 | — | |
| Versa NTSTraining tasks=8000 Tasks2022.03 | 45.16 | — | |
| MetaSGD NTSTraining tasks=4000 Tasks2022.03 | 44.88 | — | |
| Units-NTSTraining tasks=4000 Tasks2022.03 | 43.21 | — | |
| AAL-ProtoNetsLearning Paradigm=Unsupervised, Algorithm=ProtoNets2019.02 | 40.29 | — | |
| Versa NTSTraining tasks=4000 Tasks2022.03 | 39.45 | — | |
| Training from scratchLearning Paradigm=Unsupervised2019.02 | 38.48 | — | |
| Versa TSTraining tasks=4000 Tasks2022.03 | 37.86 | — | |
| MAMLTraining tasks=10000 Tasks2022.03 | 35.74 | — | |
| MAMLTraining tasks=8000 Tasks2022.03 | 33.36 | — | |
| MAMLTraining tasks=4000 Tasks2022.03 | 27.23 | — |