Skip to content

This repo explores the concept of blind source separation by training a U-Net model that separated a song into its vocal and accompaniments

License

Notifications You must be signed in to change notification settings

mohammadreza490/music-source-separation-using-Unets

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

13 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

This folder contains codes and a project report of my BSc Computing final year project. In this project, A Unet model was created to perform music source separation and separates a music into vocals and accompaniments. The implementation is based on SPECTROGRAM-CHANNELS U-NET: A SOURCE SEPARATION MODEL VIEWING EACH CHANNEL AS THE SPECTROGRAM OF EACH SOURCE

The scripts and a sample notebook have been uploaded to this repository. To access the full project folder which contains the train and test datasets, trained models and scripts, use the unzipped folder here. Make sure you download the entire Final_Year_Project folder and pass the path as the argument of Config_Handler.init() ( In order to run the scripts and train a model, you need to call the init method of Config_Handler class by passing the appropriate path that points to the unzipped folder. Have a look at the notebook in the repo for an example)

About

This repo explores the concept of blind source separation by training a U-Net model that separated a song into its vocal and accompaniments

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published