Music Genre Classification Pipeline
The pipeline is managed with python. The webAPI can be accessed with a web view.
webAPI → youtube-dl ↓ webAPI ← keras ← vamp
Current known issues: The webserver does not serve python files.
The deep learning model uses convolution on top of recurrent cells. It achieves an accuracy (precision) of 85% on three genres.
The datset is included in the form of pickled feature vectors extracted via sonic-annotator. The original music files and categorization were collections of "best-of" music CDs.
youtube-dl, sonic-annotator with plug-ins
Detailed guide in tutorial.md
On Mac OS copy the plugins to
allow execution of unsigned library in security settings after a failed attempt
sudo python3 -m pip install -r requirements.txt