5-way Few-shot Classification on miniImageNet 5-way (test)
67.761-shot AccDynamic Meta-filter (Ours)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Dynamic Meta-filter (Ours)Backbone=Res-122021.03 | 67.76 | 82.71 | |
| FEATBackbone=Res-122021.03 | 66.78 | 82.05 | |
| DeepEMDBackbone=Res-122021.03 | 65.91 | 82.41 | |
| FEATBackbone=WRN-282021.03 | 65.1 | 81.11 | |
| DSN-MRBackbone=Res-122021.03 | 64.6 | 79.51 | |
| CANBackbone=Res-122021.03 | 63.85 | 79.44 | |
| Net-CosineBackbone=Res-122021.03 | 63.85 | 81.57 | |
| E3BMBackbone=Res-122021.03 | 63.8 | 80.1 | |
| Robust dist++Backbone=WRN-282021.03 | 63.28 | 81.17 | |
| CC+rotBackbone=WRN-282021.03 | 62.93 | 79.87 | |
| MetaOptNetBackbone=Res-122021.03 | 62.64 | 78.63 | |
| MetaOptNetBackbone=ResNet-122021.06 | 62.64 | 78.63 | |
| DCBackbone=Res-122021.03 | 62.53 | 78.95 | |
| LEOBackbone=WRN-282021.03 | 61.76 | 77.59 | |
| TapNetBackbone=Res-122021.03 | 61.65 | 76.36 | |
| VFSLBackbone=Res-122021.03 | 61.23 | 77.69 | |
| wDAEBackbone=WRN-282021.03 | 61.07 | 76.75 | |
| MTLBackbone=ResNet-122021.06 | 59.84 | 77.72 | |
| PPABackbone=WRN-282021.03 | 59.6 | 73.74 | |
| TADAMBackbone=Res-122021.03 | 58.5 | 76.7 | |
| TADAMBackbone=ResNet-122021.06 | 58.5 | 76.7 | |
| Dynamic Few-shotBackbone=Conv42021.03 | 56.2 | 72.81 | |
| SNAILBackbone=Res-122021.03 | 55.71 | 68.88 | |
| VAMPIREBackbone=Standard 4-block CNN2019.07 | 51.54 | 64.31 | |
| R2D2Backbone=96-192-384-5122021.06 | 51.2 | 68.8 | |
| RelationNetBackbone=Conv42021.03 | 50.44 | 65.32 | |
| PLATIPUSBackbone=Standard 4-block CNN2019.07 | 50.13 | — | |
| ProtoNetBackbone=Conv42021.03 | 49.42 | 68.2 | |
| Prototypical NetworksBackbone=Standard 4-block CNN, Training episodes=30-way for 1-shot, 20-way for 5-shot2019.07 | 49.42 | 68.2 | |
| LLAMABackbone=Standard 4-block CNN2019.07 | 49.4 | — | |
| BMAMLBackbone=Standard 4-block CNN, Evaluation=Produced locally2019.07 | 49.17 | 64.23 | |
| MAMLBackbone=Conv42021.03 | 48.7 | 63.11 | |
| MAMLBackbone=32-32-32-322021.06 | 48.7 | 63.11 | |
| MAMLBackbone=Standard 4-block CNN2019.07 | 48.7 | 63.15 | |
| ANILBackbone=32-32-32-322021.06 | 48 | 62.2 | |
| Amortised Meta-learnerBackbone=Standard 4-block CNN2019.07 | 45 | — | |
| MatchingNetBackbone=Conv42021.03 | 43.56 | 55.31 | |
| Matching NetsBackbone=Standard 4-block CNN2019.07 | 43.56 | 55.31 | |
| Meta-LearnerBackbone=Standard 4-block CNN2019.07 | 43.44 | 60.6 |