University of Cambridge MLMI4 Team 4's replication of the paper Prototypical Networks for Few-shot Learning by Jake Snell, Kevin Swersky, Richard S. Zemel. Our implementation is inspired by this repo. We implement few-shot classification for the Omniglot and MiniImageNet datasets as well as zero-shot learning for the CUB dataset.
| Model | 1-shot (5-way Acc.) | 5-shot (5-way Acc.) | 1 -shot (20-way Acc.) | 5-shot (20-way Acc.) |
|---|---|---|---|---|
| Our replication | 98.6% | 99.6% | 95.3% | 98.7% |
| Paper | 98.8% | 99.7% | 96.0% | 98.9% |
| Model | 1-shot (5-way Acc.) | 5-shot (5-way Acc.) |
|---|---|---|
| Our replication | 48.37% | 65.37% |
| Paper | 49.42±0.78% | 68.20±0.66% |