Few-shot classification on Meta-Dataset 1.0 (test)
77.02ILSVRC AccuracyP>M>F
Evaluation results
| Method | Links | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P>M>FBackbone=ViT-base, Pre-training=DINO/IN1K, Training Source=8 in-domain datasets2022.04 | 77.02 | 91.76 | 89.73 | — | — | 80.2 | 78.28 | 95.79 | 89.86 | 64.97 | — | 84.75 | — | — | — | 92.94 | 86.94 | |
| P>M>FBackbone=ViT-base, Pre-training=DINO/IN1K, Training Source=ImageNet only2022.04 | 76.69 | 81.42 | 80.33 | — | — | 75.43 | 55.93 | 95.14 | 89.68 | 65.01 | — | 79.09 | — | — | — | 84.38 | 86.87 | |
| P>M>FBackbone=ViT-small, Pre-training=DINO/IN1K, Training Source=ImageNet only2022.04 | 74.69 | 80.68 | 76.78 | — | — | 71.25 | 54.78 | 94.57 | 88.33 | 62.57 | — | 77.53 | — | — | — | 85.04 | 86.63 | |
| P>M>FBackbone=ViT-small, Pre-training=DINO/IN1K, Training Source=8 in-domain datasets2022.04 | 74.59 | 91.79 | 88.33 | — | — | 79.23 | 74.2 | 94.12 | 88.85 | 62.59 | — | 83.13 | — | — | — | 91.02 | 86.61 | |
| P>M>FBackbone=ResNet-50, Pre-training=DINO/IN1K, Training Source=8 in-domain datasets2022.04 | 67.51 | 85.91 | 80.3 | — | — | 72.84 | 60.03 | 94.69 | 87.17 | 58.92 | — | 77.61 | — | — | — | 81.67 | 87.08 | |
| P>M>FBackbone=ResNet-50, Pre-training=DINO/IN1K, Training Source=ImageNet only2022.04 | 67.08 | 75.33 | 75.39 | — | — | 66.79 | 50.53 | 94.14 | 86.54 | 58.2 | — | 73.25 | — | — | — | 72.08 | 86.42 | |
| ProtoNetBackbone=ResNet-18, Training Source=8 in-domain datasets2022.04 | 67.01 | 44.5 | 79.56 | — | — | 65.18 | 64.88 | 40.26 | 86.85 | 46.48 | — | 63.29 | — | — | — | 71.14 | 67.01 | |
| CTXBackbone=ResNet-34, Training Source=ImageNet only2022.04 | 62.76 | 82.21 | 79.49 | — | — | 72.68 | 51.58 | 95.34 | 82.65 | 59.9 | — | 74.28 | — | — | — | 80.63 | 75.57 | |
| URLEvaluation Protocol=Varying-way Varying-shot2021.03 | 58.8 | 94.5 | 89.4 | 80.7 | 77.2 | 82.5 | 68.1 | 92 | 63.3 | 57.3 | 1.3 | — | 94.7 | 74.2 | 63.6 | — | — | |
| T-SCNAPSBackbone=ResNet-18 + Adapter, Training Source=8 in-domain datasets2022.04 | 58.8 | 93.9 | 84.1 | — | — | 78.6 | 48.8 | 91.6 | 76.1 | 48.7 | — | 72.64 | — | — | — | 76.8 | 69 | |
| Simple CNAPSEvaluation Protocol=Varying-way Varying-shot2021.03 | 58.4 | 91.6 | 82 | 74.8 | 68.8 | 76.5 | 46.6 | 90.5 | 57.2 | 48.9 | 5.2 | — | 94.6 | 74.9 | 61.3 | — | — | |
| URLBackbone=ResNet-18 + Adapter, Training Source=8 in-domain datasets2022.04 | 57.51 | 94.51 | 88.59 | — | — | 81.94 | 68.75 | 92.11 | 63.34 | 54.03 | — | 75.75 | — | — | — | 80.54 | 76.17 | |
| ITABackbone=ResNet-18 + Adapter, Training Source=8 in-domain datasets2022.04 | 57.35 | 94.96 | 89.33 | — | — | 82.01 | 67.4 | 92.18 | 83.55 | 55.75 | — | 78.07 | — | — | — | 81.42 | 76.74 | |
| SUR-merge2020.03 | 57.2 | 93.2 | 90.1 | 82.3 | 73.5 | 81.9 | 67.9 | 88.4 | 67.4 | 51.3 | — | — | 90.8 | 66.6 | 58.3 | — | — | |
| SURBackbone=ResNet-18 + Adapter, Training Source=8 in-domain datasets2022.04 | 57.2 | 93.2 | 90.1 | — | — | 81.9 | 67.9 | 88.4 | 67.4 | 51.3 | — | 75.32 | — | — | — | 82.3 | 73.5 | |
| URTEvaluation Protocol=Varying-way Varying-shot2021.03 | 56.8 | 94.2 | 85.8 | 76.2 | 71.6 | 82.4 | 64 | 87.9 | 48.3 | 51.5 | 4.4 | — | 90.6 | 67 | 57.3 | — | — | |
| SUR-pfParametric network family=true2020.03 | 56.4 | 88.5 | 79.5 | 76.4 | 73.1 | 75.7 | 48.2 | 90.6 | 65.1 | 52.1 | — | — | 93.2 | 66.4 | 57.1 | — | — | |
| SUR2020.03 | 56.3 | 93.1 | 85.4 | 71.4 | 71.5 | 81.3 | 63.1 | 82.8 | 70.4 | 52.4 | — | — | 94.3 | 66.8 | 56.6 | — | — | |
| SUREvaluation Protocol=Varying-way Varying-shot2021.03 | 56.2 | 94.1 | 85.5 | 71 | 71 | 81.8 | 64.3 | 82.9 | 51 | 52 | 5 | — | 94.3 | 66.5 | 56.9 | — | — | |
| Best SDLEvaluation Protocol=Varying-way Varying-shot2021.03 | 55.8 | 93.2 | 85.7 | 71.2 | 73 | 82.8 | 65.8 | 87 | 47.4 | 53.5 | 4.8 | — | 89.8 | 67.3 | 56.6 | — | — | |
| URTBackbone=ResNet-18 + Adapter, Training Source=8 in-domain datasets2022.04 | 55.7 | 94.4 | 85.8 | — | — | 82.5 | 63.5 | 88.2 | 69.4 | 52.2 | — | 73.98 | — | — | — | 76.3 | 71.8 | |
| BOHB-E2020.03 | 55.4 | 77.5 | 60.9 | 73.6 | 72.8 | 61.2 | 44.5 | 90.6 | 57.5 | 51.9 | — | — | — | — | — | — | — | |
| MDLEvaluation Protocol=Varying-way Varying-shot2021.03 | 53.4 | 93.8 | 86.6 | 78.6 | 71.4 | 81.5 | 61.9 | 88.7 | 51 | 49.7 | 4.6 | — | 94.4 | 66.7 | 53.6 | — | — | |
| ALFA+FP-MAMLBackbone=ResNet-12, Training Source=ImageNet only2022.04 | 52.8 | 61.87 | 63.43 | — | — | 59.17 | 41.49 | 85.96 | 60.78 | 48.11 | — | 61.41 | — | — | — | 69.75 | 70.78 | |
| CNAPs2020.03 | 52.3 | 88.4 | 80.5 | 72.2 | 58.3 | 72.5 | 47.4 | 86 | 60.2 | 42.6 | — | — | 92.7 | 61.5 | 50.1 | — | — | |
| BOHBBackbone=ResNet-18, Training Source=ImageNet only2022.04 | 51.92 | 67.57 | 54.12 | — | — | 50.33 | 41.38 | 87.34 | 51.8 | 48.03 | — | 59.15 | — | — | — | 70.69 | 68.34 | |
| BOHB-EEvaluation Protocol=Varying-way Varying-shot2021.03 | 51.9 | 67.6 | 54.1 | 70.7 | 68.3 | 50.3 | 41.4 | 87.3 | 51.8 | 48 | 8.1 | — | — | — | — | — | — | |
| FLUTEBackbone=ResNet-18, Training Source=8 in-domain datasets2022.04 | 51.8 | 93.2 | 87.2 | — | — | 79.5 | 58.1 | 91.6 | 58.4 | 50 | — | 71.78 | — | — | — | 79.2 | 68.8 | |
| CNAPSEvaluation Protocol=Varying-way Varying-shot2021.03 | 50.8 | 91.7 | 83.7 | 73.6 | 59.5 | 74.7 | 50.2 | 88.9 | 56.5 | 39.4 | 6.6 | — | — | — | — | — | — | |
| CNAPSBackbone=ResNet-18 + Adapter, Training Source=8 in-domain datasets2022.04 | 50.8 | 91.7 | 83.7 | — | — | 74.7 | 50.2 | 88.9 | 56.5 | 39.4 | — | 66.9 | — | — | — | 73.6 | 59.5 | |
| ProtoNetTraining source=ILSVRC-2012 only2019.03 | 50.5 | 59.98 | 53.1 | 68.79 | 66.56 | 48.96 | 39.71 | 85.27 | 47.12 | 41 | 2.65 | — | — | — | — | — | — | |
| ProtoNetBackbone=ResNet-18, Training Source=ImageNet only2022.04 | 50.5 | 59.98 | 53.1 | — | — | 48.96 | 39.71 | 85.27 | 47.12 | 41 | — | 56.1 | — | — | — | 68.79 | 66.56 | |
| Proto-MAMLTraining source=ILSVRC-2012 only2019.03 | 49.53 | 63.37 | 55.95 | 68.66 | 66.49 | 51.52 | 39.96 | 87.15 | 48.83 | 43.74 | 1.85 | — | — | — | — | — | — | |
| Proto-MAML2020.03 | 47.9 | 82.9 | 74.2 | 70 | 67.9 | 66.6 | 42 | 88.5 | 34.2 | 24.1 | — | — | — | — | — | — | — | |
| Proto-MAMLEvaluation Protocol=Varying-way Varying-shot2021.03 | 46.5 | 82.7 | 75.3 | 69.9 | 68.3 | 66.8 | 42 | 88.7 | 52.4 | 41.7 | 7.8 | — | — | — | — | — | — | |
| FinetuneTraining source=ILSVRC-2012 only2019.03 | 45.78 | 60.85 | 68.69 | 57.31 | 69.05 | 42.6 | 38.2 | 85.51 | 66.79 | 34.86 | 2.9 | — | — | — | — | — | — | |
| fo-MAMLTraining source=ILSVRC-2012 only2019.03 | 45.51 | 55.55 | 56.24 | 63.61 | 68.04 | 43.96 | 32.1 | 81.74 | 50.93 | 35.3 | 3.7 | — | — | — | — | — | — | |
| MatchingNetTraining source=ILSVRC-2012 only2019.03 | 45 | 52.27 | 48.97 | 62.21 | 64.15 | 42.87 | 33.97 | 80.13 | 47.8 | 34.99 | 4.65 | — | — | — | — | — | — | |
| ProtoNet2020.03 | 44.5 | 79.6 | 71.1 | 67 | 65.2 | 65.9 | 40.3 | 86.9 | 46.5 | 39.9 | — | — | — | — | — | — | — | |
| k-NNTraining source=ILSVRC-2012 only2019.03 | 41.03 | 37.07 | 46.81 | 50.13 | 66.36 | 32.06 | 36.16 | 83.1 | 44.59 | 30.38 | 5.7 | — | — | — | — | — | — | |
| RelationNetTraining source=ILSVRC-2012 only2019.03 | 34.69 | 45.35 | 40.73 | 49.51 | 52.97 | 43.3 | 30.55 | 68.76 | 33.67 | 29.15 | 6.55 | — | — | — | — | — | — | |
| MAML2020.03 | 32.4 | 71.9 | 52.8 | 47.2 | 56.7 | 50.5 | 21 | 70.9 | 34.2 | 24.1 | — | — | — | — | — | — | — | |
| FinetuneTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 2.5 | 43.08 | — | — | — | — | — | |
| FinetuneTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 3.5 | 72.03 | — | — | — | — | — | |
| fo-MAMLTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 4.5 | 37.83 | — | — | — | — | — | |
| fo-MAMLTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 1 | 76.41 | — | — | — | — | — | |
| k-NNTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 4.5 | 38.55 | — | — | — | — | — | |
| k-NNTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 7 | 64.98 | — | — | — | — | — | |
| MatchingNetTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 6 | 36.08 | — | — | — | — | — | |
| MatchingNetTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 5.5 | 69.17 | — | — | — | — | — | |
| Proto-MAMLTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 1 | 46.52 | — | — | — | — | — | |
| Proto-MAMLTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 2 | 75.23 | — | — | — | — | — | |
| ProtoNetTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 2.5 | 44.5 | — | — | — | — | — | |
| ProtoNetTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 3.5 | 71.14 | — | — | — | — | — | |
| RelationNetTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 7 | 30.89 | — | — | — | — | — | |
| RelationNetTraining source=all datasets2019.03 | — | — | — | — | — | — | — | — | — | — | 5.5 | 69.71 | — | — | — | — | — |