5-way Few-Shot Classification on MiniImageNet (1, 5, 20, 50-shot evaluation)
48.12Acc (1-shot)HMS (Ours)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| HMS (Ours)Backbone=ConvNet, Sufficient Episodic Sampling (SES)=true, Semi-normalized Similarity (SNS)=true2020.11 | 48.12 | 65.33 | 73.31 | 75.49 | |
| Baseline (Ours)Backbone=ConvNet, Sufficient Episodic Sampling (SES)=true, Semi-normalized Similarity (SNS)=true2020.11 | 47.43 | 64.11 | 72.52 | 74.72 | |
| TSP-Head (Ours)Backbone=ConvNet, Sufficient Episodic Sampling (SES)=true, Semi-normalized Similarity (SNS)=true2020.11 | 47.35 | 65.1 | 74.45 | 77.03 | |
| CUMCABackbone=ConvNet, Sufficient Episodic Sampling (SES)=true, Semi-normalized Similarity (SNS)=true2020.11 | 45.35 | 60.39 | 67.15 | 70.64 | |
| ProtoCLRBackbone=ConvNet2020.11 | 44.89 | 63.35 | 72.27 | 74.31 | |
| CACTUSBackbone=ConvNet, Sufficient Episodic Sampling (SES)=true, Semi-normalized Similarity (SNS)=true2020.11 | 44.47 | 58.24 | 65.15 | 67.08 | |
| VanillaBackbone=ConvNet2020.11 | 43.01 | 57.94 | 65.46 | 67.62 | |
| Meta-GMVAEBackbone=ConvNet2020.11 | 42.82 | 55.73 | 63.14 | 68.26 | |
| SimCLRBackbone=ConvNet2020.11 | 42.09 | 55.86 | 62.17 | 63.86 | |
| MoCo-v2Backbone=ConvNet2020.11 | 41.97 | 55 | 60.59 | 62.31 | |
| CUMCABackbone=ConvNet2020.11 | 41.12 | 54.55 | 64.45 | 70.79 | |
| MoCHiBackbone=ConvNet2020.11 | 40.81 | 55.28 | 62.89 | 65.11 | |
| ULDABackbone=ConvNet2020.11 | 40.63 | 55.41 | 63.16 | 65.2 | |
| UMTRABackbone=ConvNet2020.11 | 39.93 | 50.73 | 61.11 | 67.15 | |
| CACTUSBackbone=ConvNet2020.11 | 39.9 | 53.97 | 63.84 | 69.64 | |
| AALBackbone=ConvNet2020.11 | 37.67 | 40.29 | — | — | |
| UFLSTBackbone=ConvNet2020.11 | 33.77 | 45.03 | 53.35 | 56.72 |