Image Classification on Meta-Dataset
60.53Accuracy (ILSVRC)Transductive fine-tuning
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Transductive fine-tuningEvaluation Protocol=Transductive fine-tuning, Number of Episodes=600, Resolution=84x84, Pre-training=ImageNet-1k (ILSVRC)2019.09 | 60.53 | 82.07 | 72.4 | 82.05 | 80.47 | 57.36 | 47.72 | 92.01 | 64.37 | 42.86 | |
| Fine-tuningEvaluation Protocol=Fine-tuning, Number of Episodes=600, Resolution=84x84, Pre-training=ImageNet-1k (ILSVRC)2019.09 | 60.42 | 80.73 | 70.08 | 81.46 | 79.92 | 56.1 | 48.43 | 91.78 | 60.79 | 43.77 | |
| Support-based initializationEvaluation Protocol=Support-based initialization, Number of Episodes=600, Resolution=84x84, Pre-training=ImageNet-1k (ILSVRC)2019.09 | 60.05 | 76.32 | 61.99 | 80.16 | 74.31 | 58.75 | 48.29 | 91.23 | 53.44 | 43.84 | |
| Best in Triantafillou et al. (2019)Reference Source=Triantafillou et al. (2019), Number of Episodes=6002019.09 | 50.5 | 63.37 | 68.69 | 68.79 | 69.05 | 51.52 | 39.96 | 87.15 | 66.79 | 43.74 |