Zero-shot learning

tags

Few-shot learning, Machine learning, Transfer learning

Zero-shot learning is a type of learning task where a model has to perform prediction in an output space never seen during training. In the case of image classification, this corresponds to classifying images into classes never seen during training.