In the notebook, we define a triplet Network for use with the Omniglot dataset. We introduce the triplet loss, propose two methods of triplet selection (based on random and hard triplets), and we evaluate one-shot learning with the random and hard triplet selection methods. Furthermore, we compare these two methods against each other and discuss how one should pick a margin.
nickvdw/one-shot-learning-triplet-network
Performing one-shot learning using a triplet network with different triplet selection methods (random and hard).
Jupyter Notebook