Skip to content

QColeman97/BrahmsRestoreML

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

340 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Machine Learning Approaches to Historic Music Restoration - Brahms' 1889 Recording

Master's thesis for Cal Poly Blended Computer Science Program - Presentation

Digital signal processing (pre and post processing) is used in pair with either 2 core machine learning techniques: non-negative matrix factorization (NMF) or deep recurrent neural networks (DRNNs).

Background info, original recording (brahms.wav) & benchmark from CCRMA Webpage.

Piano samples from University of Iowa Electronic Music Studios.

Restore with NMF (best result):

python restore_with_nmf.py brahms.wav

Restore with DRNN (requires TF >= 2.0, ran on GPU = NVIDIA GTX 970):

python restore_with_drnn.py t true
python restore_with_drnn.py r true

About

Two machine learning approaches to historic music restoration, applied to the original damaged 1889 piano recording by Johannes Brahms.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages