5-way Few-Shot Image Classification on tieredImageNet (test)
77.71-shot AccLST
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LSTBackbone=ResNet-12 (pre), Strategy=recursive, hard, soft2019.06 | 77.7 | 85.2 | |
| AlignSetting=Supervised, Backbone=WRN-28-102021.05 | 74.4 | 86.61 | |
| Sum-minBackbone=SF122022.04 | 73.63 | 87.59 | |
| Transductive fine-tuningBackbone=WRN-28-10, Pre-training split=train2019.09 | 73.34 | 85.5 | |
| Transductive fine-tuningArchitecture=WRN-28-10, Pre-training Split=train2019.09 | 73.34 | 85.5 | |
| Transductive fine-tuningBackbone=WRN-28-10, Pre-training split=train + val2019.09 | 72.87 | 86.15 | |
| Transductive fine-tuningArchitecture=WRN-28-10, Pre-training Split=train + val2019.09 | 72.87 | 86.15 | |
| Distill [45]Backbone=ResNet122022.04 | 72.21 | 87.08 | |
| MELRBackbone=ResNet122022.04 | 72.14 | 87.01 | |
| Dynamic Meta-filter (Ours)Backbone=Res-122021.03 | 71.89 | 85.96 | |
| DMFBackbone=ResNet122022.04 | 71.89 | 85.96 | |
| Min-minBackbone=SF122022.04 | 71.75 | 86.4 | |
| DistillSetting=Supervised, Backbone=ResNet-122021.05 | 71.52 | 86.03 | |
| Distill [55]Backbone=ResNet122022.04 | 71.52 | 86.03 | |
| SimpleShot (CL2N)Network=DenseNet2019.11 | 71.32 | 86.66 | |
| Match-sumBackbone=SF122022.04 | 71.22 | 85.43 | |
| E3BMBackbone=Res-122021.03 | 71.2 | 85.3 | |
| DeepEMDBackbone=Res-122021.03 | 71.16 | 86.03 | |
| DeepEMDBackbone=ResNet122022.04 | 71.16 | 86.03 | |
| MixtFSLBackbone=ResNet122022.04 | 70.97 | 86.16 | |
| FEATBackbone=Res-122021.03 | 70.8 | 84.79 | |
| FEATBackbone=ResNet122022.04 | 70.8 | 84.79 | |
| CC+rotBackbone=WRN-282021.03 | 70.53 | 84.98 | |
| CC+rotSetting=Supervised, Backbone=WRN-28-102021.05 | 70.53 | 84.98 | |
| FEATBackbone=WRN-282021.03 | 70.41 | 84.38 | |
| SimpleShot (L2N)Network=DenseNet2019.11 | 69.91 | 86.42 | |
| CANBackbone=Res-122021.03 | 69.89 | 84.23 | |
| VAMPIRE2019.07 | 69.87 | 82.7 | |
| SimpleShot (CL2N)Network=WRN2019.11 | 69.75 | 85.31 | |
| SimpleBackbone=ResNet122022.04 | 69.74 | 84.41 | |
| Transductive fine-tuningBackbone=ResNet-12, Pre-training split=train + val2019.09 | 69.48 | 84.37 | |
| SimpleShot (CL2N)Network=MobileNet2019.11 | 69.47 | 85.17 | |
| SimpleShot (CL2N)Network=ResNet-182019.11 | 69.09 | 84.58 | |
| PDA-NetSetting=Unsupervised, Backbone=ResNet-502021.05 | 69.01 | 84.2 | |
| RELATIONNET2 (DCN)Embedding=SENet2018.11 | 68.83 | 79.62 | |
| SimpleShot (L2N)Network=MobileNet2019.11 | 68.66 | 85.43 | |
| SimpleShot (L2N)Network=ResNet-182019.11 | 68.64 | 84.47 | |
| RELATIONNET2 (DCN)Embedding=ResNet-122018.11 | 68.58 | 80.65 | |
| Transductive fine-tuningBackbone=ResNet-12, Pre-training split=train2019.09 | 68.41 | 83.41 | |
| ProtoNet+Backbone=ResNet122022.04 | 68.23 | 84.03 | |
| wDAEBackbone=WRN-282021.03 | 68.18 | 83.09 | |
| Support-based initializationBackbone=ResNet-12, Pre-training split=train + val2019.09 | 67.6 | 83.09 | |
| Support-based initializationBackbone=WRN-28-10, Pre-training split=train2019.09 | 67.45 | 82.88 | |
| Support-based initializationArchitecture=WRN-28-10, Pre-training Split=train2019.09 | 67.45 | 82.88 | |
| DSN-MRBackbone=Res-122021.03 | 67.39 | 82.85 | |
| Support-based initializationBackbone=WRN-28-10, Pre-training split=train + val2019.09 | 67.34 | 83.32 | |
| Support-based initializationArchitecture=WRN-28-10, Pre-training Split=train + val2019.09 | 67.34 | 83.32 | |
| MTLBackbone=ResNet-122021.06 | 67.11 | 83.69 | |
| SimpleShot (L2N)Network=WRN2019.11 | 66.86 | 85.5 | |
| Fine-tuningBackbone=WRN-28-10, Pre-training split=train2019.09 | 66.58 | 85.55 | |
| Fine-tuningArchitecture=WRN-28-10, Pre-training Split=train2019.09 | 66.58 | 85.55 | |
| FinetuneSetting=Supervised, Backbone=WRN-28-102021.05 | 66.58 | 85.55 | |
| Support-based initializationBackbone=ResNet-12, Pre-training split=train2019.09 | 66.39 | 81.93 | |
| LEOBackbone=WRN-28-10 (pre), Method Category=Gradient descent2019.06 | 66.33 | 81.44 | |
| LEONetwork=WRN2019.11 | 66.33 | 81.44 | |
| LEOBackbone=WRN-28-10, Pre-training split=train + val2019.09 | 66.33 | 81.44 | |
| LEOArchitecture=WRN-28-10, Pre-training Split=train + val2019.09 | 66.33 | 81.44 | |
| LEOBackbone=WRN-282021.03 | 66.33 | 81.44 | |
| LEOSetting=Supervised, Backbone=WRN-28-102021.05 | 66.33 | 81.44 | |
| LEO2019.07 | 66.33 | 81.44 | |
| Fine-tuningBackbone=WRN-28-10, Pre-training split=train + val2019.09 | 66.23 | 86.08 | |
| Fine-tuningArchitecture=WRN-28-10, Pre-training Split=train + val2019.09 | 66.23 | 86.08 | |
| DNSBackbone=ResNet122022.04 | 66.22 | 82.79 | |
| MetaOpt-SVMBackbone=ResNet-12, Method Category=Gradient descent, Note=Using 15-shot training samples on every meta-train task2019.06 | 65.99 | 81.56 | |
| MetaOpt SVMBackbone=ResNet-12*, Pre-training split=train2019.09 | 65.99 | 81.56 | |
| MetaOpt SVMArchitecture=ResNet-12, Pre-training Split=train2019.09 | 65.99 | 81.56 | |
| MetaOptNetBackbone=Res-122021.03 | 65.99 | 81.56 | |
| MetaOptNetSetting=Supervised, Backbone=ResNet-122021.05 | 65.99 | 81.56 | |
| MetaOptNetBackbone=ResNet-122021.06 | 65.99 | 81.56 | |
| OptNetBackbone=ResNet122022.04 | 65.99 | 81.56 | |
| METAOPTNETEmbedding=ResNet-12, Wider ResNet=true, Gradient-based optimization at meta-test time=true2018.11 | 65.81 | 81.75 | |
| MetaOpt SVMBackbone=ResNet-12*, Pre-training split=train + val2019.09 | 65.81 | 81.75 | |
| MetaOpt SVMArchitecture=ResNet-12*, Pre-training Split=train + val2019.09 | 65.81 | 81.75 | |
| MetaOptNet2019.07 | 65.81 | 81.75 | |
| MTLBackbone=ResNet122022.04 | 65.62 | 80.61 | |
| MTLBackbone=ResNet-12 (pre), Method Category=Gradient descent, Reported By=Current Authors2019.06 | 65.6 | 78.6 | |
| Fine-tuningBackbone=ResNet-12, Pre-training split=train + val2019.09 | 65.55 | 84.51 | |
| SimpleShot (CL2N)Network=ResNet-102019.11 | 65.37 | 81.84 | |
| SimpleShot (L2N)Network=ResNet-102019.11 | 64.58 | 82.31 | |
| Fine-tuningBackbone=ResNet-12, Pre-training split=train2019.09 | 64.45 | 83.59 | |
| SimpleShot (UN)Network=DenseNet2019.11 | 64.35 | 85.69 | |
| SimpleShot (UN)Network=WRN2019.11 | 63.85 | 84.17 | |
| SimpleShot (UN)Network=MobileNet2019.11 | 63.65 | 84.01 | |
| TAP NETEmbedding=ResNet-122018.11 | 63.08 | 80.26 | |
| TapNetBackbone=ResNet122022.04 | 63.08 | 80.26 | |
| Meta SGDNetwork=WRN2019.11 | 62.95 | 79.34 | |
| SimpleShot (UN)Network=ResNet-182019.11 | 62.69 | 83.27 | |
| RELATIONNET2 (DCN)Embedding=Conv-42018.11 | 60.58 | 72.42 | |
| TPNEmbedding=Conv-4, Semi-supervised or transductive=true2018.11 | 59.91 | 73.3 | |
| Transductive PropagationBackbone=conv (64)x4, Pre-training split=train2019.09 | 59.91 | 73.3 | |
| Transductive PropagationArchitecture=conv (64)x42019.09 | 59.91 | 73.3 | |
| Support-based initializationBackbone=conv (64)x4, Pre-training split=train + val2019.09 | 59.08 | 74.62 | |
| Transductive fine-tuningBackbone=conv (64)x4, Pre-training split=train + val2019.09 | 58.81 | 74.71 | |
| SimpleShot (UN)Network=ResNet-102019.11 | 58.6 | 79.99 | |
| Support-based initializationBackbone=conv (64)x4, Pre-training split=train2019.09 | 58.42 | 73.98 | |
| Transductive fine-tuningBackbone=conv (64)x4, Pre-training split=train2019.09 | 58.05 | 74.24 | |
| Fine-tuningBackbone=conv (64)x4, Pre-training split=train + val2019.09 | 58.04 | 74.48 | |
| Fine-tuningBackbone=conv (64)x4, Pre-training split=train2019.09 | 57.45 | 73.96 | |
| Transductive Propagation Networks2019.07 | 57.41 | 71.55 | |
| RELATION NETEmbedding=Conv-42018.11 | 54.48 | 71.31 |