Skip to main content
. 2021 Jul 21;7:e613. doi: 10.7717/peerj-cs.613

Table 3. Small sample classification experimental results in Omniglot dataset.

Model Fine-tuning Accuracy
5-way 5-shot 20-way 5-shot
MAML Y 99.89 ± 0.10% 98.90 ± 0.13%
Meta-SGD Y 99.91 ± 0.11% 98.92 ± 0.22%
Multi-scale relational networks N 99.70 ± 0.08% 99.01 ± 0.13%
Multi-scale meta-relational network N 99.84 ± 0.16% 99.20 ± 0.12%