PANNs

PANNs (Pretrained Audio Neural Networks) are a family of pretrained audio neural networks trained on the large-scale AudioSet dataset, for audio pattern recognition tasks like audio tagging and sound event detection.

The idea is the same as pretrained models in computer vision: train once on a big, general dataset, then Transfer Learning the model to other audio tasks.

From paper PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognition.