Hands-On One-shot Learning with Python

Data preprocessing

In a general deep learning setting, to train a model on a given dataset, D, we divide our dataset into three parts—training, validation, and test set. But in the meta-learning setting, we first divide the dataset into task-specific sets (for example, cat breed classification and dog breed classification) known as meta sets, say . For each D consists of and , so for K-shot learning, each consists of K*N examples, where N is ...