The aim of the project is the classification of music genres from audio files.
The dataset used is the following: Click here
We have pre-processed the dataset and you can find it here
This folder contains the trained neural network models.
This folder contains the report, written in latex, which describes the project and the tasks performed.
The notebooks folder contains the .ipynb files while the project contains the python project with the utility functions and features extraction and processing.
"CNN_audio.ipynb" is used to train a CNN like the one described in the paper by "Daniel Kostrzewa".
"CNN_tuning.ipynb" is used to tune, with Hyperband, a CNN.
"analysis.ipynb" contains the EDA for the raw audio file and the mel spectrograms.
"handcraft_nn.ipynb" is used to train and evaluate the neural network for the handcrafted features.
"features_extraction.ipynb" is used to extract features from a specific layer and evaluate using different classifiers.