ronggong / Singing-Voice-Separation

Singing-Voice-Separation

This is an implementation of U-Net for vocal separation with tensorflow

Requirement

• librosa==0.6.2
• numpy==1.14.3
• tensorflow==1.9.0
• python==3.6.5

Data

I prepare CCMixter datasets in "./data" and Each track consisted of Mixed, instrumental, Vocal version

$ python CCMixter_process.py

Usage

• Train

$ python Training.py

• Test

$ python Test.py

Paper

Andreas Jansson, et al. SINGING VOICE SEPARATION WITH DEEP U-NET CONVOLUTIONAL NETWORKS. 2017.
paper: https://ismir2017.smcnus.org/wp-content/uploads/2017/10/171_Paper.pdf