Welcome to my degree project about music genre prediction using Neural Networks with Music Information Retrieval and the GTZAN dataset.
To run this project you need to execute this steps (in the sequence):
- Install the requirements
- Download the GTZAN dataset from http://marsyas.info/downloads/datasets.html
- Put the GTZAN "genres" folder inside the root of current project
- Run the file "create_datasets.py" that creates the datasets with music features (MFCCs)
- Put some songs that you want to predict the genre inside the "songs" folder
- Run the file
predict_genre.py
and enjoy !
To customize the network models:
The trained network models are the ".h5" files, and if you want to change some model, it is only necessary to change the correspondent classifier.py
file and run it (e.g., cnn_genre_classifier.py
). After run it, it will save your new network model generating a new ".h5" and after that you can enjoy your custom model running the predict_genre.py
.