Repo for code used for master thesis at Warsaw University of Technology. Thesis included denoising and upsampling audio files using DNN architectures such as autoencoders, SEGAN and WaveNet.
To setup on ubuntu, run:
./scripts/setup_project.sh
pip install -r requirements-to-freeze.txt
Not sure if it will work on other systems.
Demo app allows to:
- load audio,
- apply selected noise,
- process audio with NN model
- play original audio or any audio generated by applying noise or model inference
- save audio generated by applying noise or model inference
To run demo:
python src/demo.py