Tensorflow 2.0 implementation of the paper A Fully Convolutional Neural Network for Speech Enhancement
Blog post: Practical Deep Learning Audio Denoising
Part of the dataset used to train the original system is now available to download
The zip file contains 1 training file (that is 10% of the data used to train the system), a validation file, and two audio files (not included in the training files) used to evaluate the model.
You can create the dataset for yourself.
- Download the Mozilla Common Voice and UrbanSound8K datasets.
- Use the
create_dataset.pyscript to create the TFRecord files.