Few-shot classification on CUB-200 2011 (test)
84.365-way 1-shot AccFRN+ + TDM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FRN+ + TDMBackbone=ResNet-122022.07 | 84.36 | 93.37 | — | |
| FRN+Backbone=ResNet-122022.07 | 83.54 | 92.96 | — | |
| CTX+ + TDMBackbone=ResNet-122022.07 | 83.45 | 92.49 | — | |
| FCAMBackbone=ResNet122024.07 | 82.89 | 93.06 | — | |
| ADRGNBackbone=ResNet12, Transductive=true, Attribute Information=true2024.07 | 82.32 | 92.97 | — | |
| CausalFSFGBackbone=Conv-42025.12 | 81.94 | 93.33 | — | |
| CTX+Backbone=ResNet-122022.07 | 80.95 | 91.54 | — | |
| DSN+ + TDMBackbone=ResNet-122022.07 | 80.58 | 89.95 | — | |
| DSN+Backbone=ResNet-122022.07 | 80.47 | 89.92 | — | |
| AGAMBackbone=ResNet12, Attribute Information=true2024.07 | 79.58 | 87.17 | — | |
| MCLBackbone=ResNet12, Data Split=Reproduced by current paper2024.07 | 79.39 | 90.48 | — | |
| DeepEMD v2Backbone=ResNet122024.07 | 79.27 | 89.8 | — | |
| ProtoNet + TDMBackbone=ResNet-122022.07 | 79.11 | 90.83 | — | |
| Bi-FRNBackbone=Conv-4, Published In=AAAI 20232025.12 | 79.08 | 92.22 | — | |
| FRNBackbone=ResNet12, Data Split=Reproduced by current paper2024.07 | 78.86 | 90.48 | — | |
| ProtoNetBackbone=ResNet-122022.07 | 78.58 | 89.83 | — | |
| Oblique ManifoldBackbone=ResNet182024.07 | 78.24 | 92.15 | — | |
| ECKPNBackbone=ResNet12, Transductive=true2024.07 | 77.43 | 92.21 | — | |
| SSL-FEW-SHOTEmbedding Net=AmdimNet, Training Mode=Self-Supervised Learning (Image1K_SSL), Meta-learning=true2019.11 | 77.09 | 89.18 | — | |
| DPGNbackbone=ConvNet, transduction=true2020.03 | 76.05 | 89.08 | — | |
| DPGNbackbone=ResNet12, transduction=true2020.03 | 75.71 | 91.48 | — | |
| DeepEMDBackbone=ResNet122024.07 | 75.65 | 88.69 | — | |
| FRNBackbone=Conv-4, Published In=CVPR 20212025.12 | 74.9 | 89.39 | — | |
| PARNBackbone=Conv-4, Published In=ICCV 20192025.12 | 74.43 | 83.11 | — | |
| Afrasiyabi et al.Backbone=ResNet-182022.07 | 74.22 | 88.65 | — | |
| Centroid AlignmentBackbone=ResNet182024.07 | 74.22 | 88.65 | — | |
| MatchNetBackbone=ResNet-182022.07 | 73.42 | 84.45 | — | |
| ProtoNetBackbone=ResNet-182022.07 | 72.99 | 86.65 | — | |
| Neg-CosineBackbone=ResNet-182022.07 | 72.66 | 89.4 | — | |
| CTXBackbone=Conv-4, Published In=NeurIPS 20202025.12 | 72.61 | 86.23 | — | |
| FRN+TDMBackbone=Conv-4, Published In=CVPR 20222025.12 | 72.01 | 89.05 | — | |
| MatchNetBackbone=ResNet122024.07 | 71.87 | 85.08 | — | |
| SSL-FEW-SHOTEmbedding Net=AmdimNet, Training Mode=Self-Supervised Learning (CUB150_SSL), Meta-learning=true2019.11 | 71.85 | 84.29 | — | |
| S2M2Backbone=ResNet-182022.07 | 71.43 | 85.55 | — | |
| FEATbackbone=ResNet122020.03 | 68.87 | 82.9 | — | |
| RelatioNetBackbone=ResNet-182022.07 | 68.58 | 84.05 | — | |
| MAMLBackbone=ResNet-182022.07 | 68.42 | 83.47 | — | |
| Baseline++Backbone=ResNet-182022.07 | 67.02 | 83.58 | — | |
| ProtoNetBackbone=ResNet122024.07 | 66.09 | 82.5 | — | |
| BaselineBackbone=ResNet-182022.07 | 65.51 | 82.85 | — | |
| SAMLBackbone=Conv-4, Published In=ICCV 20192025.12 | 65.35 | 78.47 | — | |
| ProtoNetBackbone=Conv-4, Published In=NeurIPS 20172025.12 | 64.82 | 85.74 | — | |
| DeepEMDBackbone=Conv-4, Published In=CVPR 20202025.12 | 64.08 | 80.55 | — | |
| RelationBackbone=Conv-4, Published In=CVPR 20182025.12 | 63.94 | 77.87 | — | |
| LRPABNBackbone=Conv-4, Published In=TMM 20212025.12 | 63.63 | 76.06 | — | |
| BSNet(D&C)Backbone=Conv-4, Published In=TIP 20212025.12 | 62.84 | 85.39 | — | |
| RelationNetbackbone=ConvNet2020.03 | 62.45 | 76.11 | — | |
| RelationNetEmbedding Net=4 Conv, Meta-learning=true2019.11 | 62.45 | 76.11 | — | |
| MatchingNetbackbone=ConvNet2020.03 | 61.16 | 72.86 | — | |
| MatchingNetEmbedding Net=4 Conv, Meta-learning=true2019.11 | 61.16 | 72.86 | — | |
| MACOEmbedding Net=4 Conv, Meta-learning=true2019.11 | 60.76 | 74.96 | — | |
| CloserLookbackbone=ConvNet2020.03 | 60.53 | 79.34 | — | |
| Baseline++Embedding Net=4 Conv, Meta-learning=true2019.11 | 60.53 | 79.34 | — | |
| RényiCLmulti-crop data augmentation=false, number of episodes=2000, harder augmentations=true2022.08 | 59.73 | 82.12 | — | |
| RényiCLmulti-crop data augmentation=true, number of episodes=2000, harder augmentations=true2022.08 | 58.25 | 82.38 | — | |
| DN4Backbone=Conv-4, Published In=CVPR 20192025.12 | 57.45 | 84.41 | — | |
| MAMLbackbone=ConvNet2020.03 | 55.92 | 72.09 | — | |
| MAMLEmbedding Net=4 Conv, Meta-learning=true2019.11 | 55.92 | 72.09 | — | |
| InfoMinmulti-crop data augmentation=false, number of episodes=2000, harder augmentations=true2022.08 | 53.81 | 72.2 | — | |
| DN4backbone=ConvNet2020.03 | 53.15 | 81.9 | — | |
| DN4-DAEmbedding Net=4 Conv, Meta-learning=true2019.11 | 53.15 | 81.9 | — | |
| CLSAmulti-crop data augmentation=true, number of episodes=2000, harder augmentations=true2022.08 | 52.87 | 70.92 | — | |
| ProtoNetbackbone=ConvNet2020.03 | 51.31 | 70.77 | — | |
| ProtoNetEmbedding Net=4 Conv, Meta-learning=true2019.11 | 51.31 | 70.77 | — | |
| CLSAmulti-crop data augmentation=false, number of episodes=2000, harder augmentations=true2022.08 | 50.83 | 67.93 | — | |
| ArLBackbone=ConvNet2022.04 | 50.6 | 65.9 | — | |
| SSL-FEW-SHOTEmbedding Net=AmdimNet, Training Mode=Supervised Learning (CUB150_SL), Meta-learning=true2019.11 | 45.1 | 74.59 | — | |
| SSL-FEW-SHOTEmbedding Net=AmdimNet, Training Mode=Self-Supervised Learning (CUB150_SSL), Meta-learning=false2019.11 | 40.83 | 65.27 | — | |
| CLIPNumber of ways=5, Number of shots=12023.10 | — | — | 75.21 | |
| CLIPNumber of ways=5, Number of shots=52023.10 | — | — | 91.48 | |
| EPNetbackbone=RESNET-12, shots=1-shot, resolution=224 x 2242020.03 | — | — | 82.85 | |
| EPNetbackbone=RESNET-12, shots=5-shot, resolution=224 x 2242020.03 | — | — | 91.32 | |
| EPNetbackbone=WRN-28-10, shots=1-shot, resolution=224 x 2242020.03 | — | — | 87.75 | |
| EPNetbackbone=WRN-28-10, shots=5-shot, resolution=224 x 2242020.03 | — | — | 94.03 | |
| FD-AlignNumber of ways=5, Number of shots=12023.10 | — | — | 82.38 | |
| FD-AlignNumber of ways=5, Number of shots=52023.10 | — | — | 93.87 | |
| Manifold mixupbackbone=WRN-28-10, shots=1-shot, resolution=224 x 2242020.03 | — | — | 80.68 | |
| Manifold mixupbackbone=WRN-28-10, shots=5-shot, resolution=224 x 2242020.03 | — | — | 90.85 | |
| Robust-20++backbone=RESNET-18, shots=1-shot2020.03 | — | — | 68.68 | |
| Robust-20++backbone=RESNET-18, shots=5-shot2020.03 | — | — | 83.21 | |
| WiSE-FTNumber of ways=5, Number of shots=12023.10 | — | — | 81.16 | |
| WiSE-FTNumber of ways=5, Number of shots=52023.10 | — | — | 93.41 |