Few-Shot Image Classification on mini-ImageNet (test) (1, 5, 10-shot)
83.35Acc (5-shot)S2M2R + Graph (Ours)
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| S2M2R + Graph (Ours)Backbone=WRN 28-102021.02 | 83.35 | 66.93 | — | — | — | — | |
| S2M2_RBackbone=WRN-28-10, N-way=5-way2019.07 | 83.18 | 64.93 | — | — | — | — | |
| S2M2RBackbone=WRN 28-102021.02 | 83.18 | 64.93 | — | — | — | — | |
| SimpleShotBackbone=WRN 28-102021.02 | 82.09 | 65.87 | — | — | — | — | |
| Robust20-distAggregation=SUR, Backbone=ResNet-122020.03 | 81.19 | 63.93 | — | — | — | — | |
| KGTNBackbone=WRN 28-102021.02 | 81.07 | 65.71 | — | — | — | — | |
| Robust20-distAggregation=last, Backbone=ResNet-122020.03 | 81.06 | 64.14 | — | — | — | — | |
| RotationBackbone=WRN-28-10, N-way=5-way2019.07 | 81.03 | 63.9 | — | — | — | — | |
| Robust20-distAggregation=concat, Backbone=ResNet-122020.03 | 80.79 | 63.22 | — | — | — | — | |
| S2M2_RBackbone=ResNet-18, N-way=5-way2019.07 | 80.58 | 64.06 | — | — | — | — | |
| DenseClsAggregation=SUR, Backbone=ResNet-122020.03 | 80.04 | 63.13 | — | — | — | — | |
| ProtoNetBackbone=WRN 28-102021.02 | 79.97 | 62.6 | — | — | — | — | |
| DenseClsAggregation=concat, Backbone=ResNet-122020.03 | 79.59 | 62.74 | — | — | — | — | |
| S2M2_RBackbone=ResNet-34, N-way=5-way2019.07 | 79.45 | 63.74 | — | — | — | — | |
| S2M2_EBackbone=WRN-28-10, N-way=5-way2019.07 | 79.35 | 62.33 | — | — | — | — | |
| ClsAggregation=SUR, Backbone=ResNet-122020.03 | 79.25 | 60.79 | — | — | — | — | |
| ExemplarBackbone=WRN-28-10, N-way=5-way2019.07 | 78.8 | 62.2 | — | — | — | — | |
| Baseline++Backbone=WRN 28-102021.02 | 78.8 | 59.62 | — | — | — | — | |
| DCOBackbone=WRN-28-10, N-way=5-way2019.07 | 78.63 | 62.64 | — | — | — | — | |
| DenseClsAggregation=last, Backbone=ResNet-122020.03 | 78.25 | 62.61 | — | — | — | — | |
| LEOBackbone=WRN-28-10, N-way=5-way2019.07 | 77.59 | 61.76 | — | — | — | — | |
| LEO (train+val)Backbone=WRN 28-102021.02 | 77.59 | 61.76 | — | — | — | — | |
| Mixup (a = 1)Backbone=WRN-28-10, N-way=5-way2019.07 | 77.52 | 59.65 | — | — | — | — | |
| TADAMClassification Way=5-way, Backbone=ResNet-122018.05 | 76.7 | 58.5 | 80.8 | — | — | — | |
| MatchingNetBackbone=WRN 28-102021.02 | 76.32 | 64.03 | — | — | — | — | |
| EGNNprotocol=5-way2020.08 | 76.3 | 44.74 | 77.4 | — | 75.49 | 72.83 | |
| ClsAggregation=last, Backbone=ResNet-122020.03 | 76.28 | 60.09 | — | — | — | — | |
| Manifold MixupBackbone=WRN-28-10, N-way=5-way2019.07 | 75.89 | 57.16 | — | — | — | — | |
| ClsAggregation=concat, Backbone=ResNet-122020.03 | 75.67 | 57.15 | — | — | — | — | |
| Discriminative k-shotClassification Way=5-way2018.05 | 73.9 | 56.3 | 78.5 | — | — | — | |
| QiaoBackbone=WRN 28-102021.02 | 73.74 | 59.6 | — | — | — | — | |
| ProtoNetBackbone=WRN-28-10, N-way=5-way2019.07 | 73.68 | 54.16 | — | — | — | — | |
| Baseline++Backbone=WRN-28-10, N-way=5-way2019.07 | 72.99 | 57.53 | — | — | — | — | |
| MELRBackbone channels=64, Number of ways=52022.01 | 72.3 | 55.4 | — | — | — | — | |
| GCRBackbone channels=64, Number of ways=52022.01 | 72.3 | 53.2 | — | — | — | — | |
| TIMBackbone=WRN28-102024.12 | 72.1 | — | 74.9 | — | 76.2 | — | |
| adaResNetClassification Way=5-way2018.05 | 71.9 | 56.9 | — | — | — | — | |
| PARNBackbone channels=64, Number of ways=52022.01 | 71.6 | 55.2 | — | — | — | — | |
| UNEM-GaussianBackbone=WRN28-102024.12 | 71.6 | — | 79.2 | — | 83.7 | — | |
| α-TIMBackbone=WRN28-102024.12 | 71.5 | — | 75.2 | — | 78.3 | — | |
| KTNBackbone channels=64, Number of ways=52022.01 | 71.2 | 54.6 | — | — | — | — | |
| RelationNetBackbone=WRN-28-10, N-way=5-way2019.07 | 70.2 | 52.19 | — | — | — | — | |
| CSSL-FSL_Image168protocol=5-way, pre-training=168 classes from ImageNet2020.08 | 68.91 | 54.17 | 74.82 | — | 78.47 | 80.83 | |
| SNAILClassification Way=5-way2018.05 | 68.9 | 55.7 | — | — | — | — | |
| HyperTransformerBackbone channels=64, Number of ways=52022.01 | 68.5 | 54.1 | — | — | — | — | |
| Proto NetsClassification Way=5-way, re-implementation=true2018.05 | 68.2 | 49.4 | 74.3 | — | — | — | |
| PNBackbone channels=64, Number of ways=52022.01 | 68.2 | 49.4 | — | — | — | — | |
| α-AMBackbone=WRN28-102024.12 | 68.2 | — | 71.3 | — | 73.3 | — | |
| HyperTransformer-48Backbone channels=48, Number of ways=52022.01 | 68.1 | 55.1 | — | — | — | — | |
| PPABackbone channels=64, Number of ways=52022.01 | 67.9 | 54.5 | — | — | — | — | |
| TIMBackbone=ResNet-182024.12 | 66.8 | — | 69.9 | — | 70.8 | — | |
| α-TIMBackbone=ResNet-182024.12 | 66.7 | — | 71 | — | 73.9 | — | |
| MAMLBackbone=WRN-28-10, N-way=5-way2019.07 | 66.62 | 54.69 | — | — | — | — | |
| SAMLBackbone channels=64, Number of ways=52022.01 | 66.5 | 52.2 | — | — | — | — | |
| UNEM-GaussianBackbone=ResNet-182024.12 | 66.4 | — | 75.6 | — | 80.4 | — | |
| TAMLBackbone channels=64, Number of ways=52022.01 | 66.1 | 51.8 | — | — | — | — | |
| Relation NetClassification Way=5-way2018.05 | 65.3 | 50.4 | — | — | — | — | |
| IMPBackbone channels=64, Number of ways=52022.01 | 64.7 | 49.2 | — | — | — | — | |
| α-AMBackbone=ResNet-182024.12 | 64.4 | — | 67.8 | — | 70.1 | — | |
| MAMLClassification Way=5-way2018.05 | 63.1 | 48.7 | — | — | — | — | |
| PADDLEBackbone=ResNet-182024.12 | 62.9 | — | 73.5 | — | 79.8 | — | |
| PADDLEBackbone=WRN28-102024.12 | 62.6 | — | 73 | — | 79.2 | — | |
| LaplacianShotBackbone=WRN28-102024.12 | 61 | — | 66.8 | — | 71 | — | |
| Meta NetsClassification Way=5-way2018.05 | 60.6 | 43.4 | — | — | — | — | |
| Matching NetworksClassification Way=5-way2018.05 | 60 | 46.6 | — | — | — | — | |
| BaselineBackbone=WRN28-102024.12 | 59 | — | 65.7 | — | 72.1 | — | |
| LR+ICIBackbone=WRN28-102024.12 | 58.8 | — | 65.7 | — | 72 | — | |
| LaplacianShotBackbone=ResNet-182024.12 | 57.9 | — | 64.2 | — | 68.3 | — | |
| BaselineBackbone=ResNet-182024.12 | 55.4 | — | 62.1 | — | 67.9 | — | |
| LR+ICIBackbone=ResNet-182024.12 | 55.4 | — | 62.1 | — | 68.1 | — | |
| MNBackbone channels=64, Number of ways=52022.01 | 55.3 | 43.6 | — | — | — | — | |
| BD-CSPNBackbone=WRN28-102024.12 | 51.1 | — | 55.5 | — | 58.4 | — | |
| BD-CSPNBackbone=ResNet-182024.12 | 49.8 | — | 54.6 | — | 56.5 | — | |
| PT-MAPBackbone=WRN28-102024.12 | 26.5 | — | 28 | — | 29.3 | — | |
| PT-MAPBackbone=ResNet-182024.12 | 25.7 | — | 27.2 | — | 28.4 | — | |
| Cosine + AttentionInput size=224, Network=ResNet2019.03 | — | — | — | 56.2 | — | — | |
| Cosine ClassifierInput size=224, Network=ResNet2019.03 | — | — | — | 51.87 | — | — | |
| FEATInput size=80, Network=WideResNet2019.03 | — | — | — | 61.72 | — | — | |
| LEOInput size=80, Network=WideResNet2019.03 | — | — | — | 61.76 | — | — | |
| Linear ClassifierInput size=224, Network=ResNet2019.03 | — | — | — | 51.75 | — | — | |
| PPAInput size=80, Network=WideResNet2019.03 | — | — | — | 59.6 | — | — | |
| Robust 20 FullInput size=224, Network=ResNet2019.03 | — | — | — | 63.95 | — | — | |
| Robust 20 FullInput size=84, Network=ResNet2019.03 | — | — | — | 59.38 | — | — | |
| Robust 20 FullInput size=80, Network=WideResNet2019.03 | — | — | — | 63.46 | — | — | |
| Robust 20-dist++Input size=224, Network=ResNet2019.03 | — | — | — | 63.73 | — | — | |
| Robust 20-dist++Input size=84, Network=ResNet2019.03 | — | — | — | 59.48 | — | — | |
| Robust 20-dist++Input size=80, Network=WideResNet2019.03 | — | — | — | 63.28 | — | — | |
| TADAMInput size=84, Network=ResNet2019.03 | — | — | — | 58.5 | — | — |