Summary

Semi-supervised few-shot learning for sound event detection. Addresses challenge of annotating large audio datasets. Uses prototypical networks and meta-learning to learn from few labeled examples plus unlabeled audio segments.