A Fully Convolutional Neural Network for Speech Enhancement

Tensorflow 2.0 implementation of the paper A Fully Convolutional Neural Network for Speech Enhancement

Blog post: Practical Deep Learning Audio Denoising

Dataset

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.py script to create the TFRecord files.