- This project uses Support Vector Machines, Neural Networks, and Convolutional Neural networks to mp3 files into one of 6 genres. This is done by first processing the train mp3 datasets with SVD, MFCC and a Spectrograms. The processed data is then used to train each of the tree classifiers. The twsting datasets are also processed the same way as the training datasets and are used to predict the genres of the mp3 files.
To install this project, you can clone the git repo to your local machine or, in GitHub, under the Code tab, click on the Green Code button, and the choose Download Zip.
The testing files and training files must be downloaded from: https://www.kaggle.com/competitions/cs529-project-3-audio/data.
Files are too large to include in our repo but must be placed in the same working directory as the rest of the project. The folder must be extracted in such a way that the train and test folders are in the same working directory as the rest of the code.
For running SVM from the command line in a Linux terminal:
-
cd into the directory where "Svm.py", "train.csv", "train" directory, "test" directory, and "test_idx.csv" are located (these should all be in the same directory with the exact names listed).
-
Once in the correct directory, simply type
python Svm.pyinto the terminal. -
For Windows, the process is the same, however to run, you will need to enter
py Svm.pyand follow the prompts outlined as above. -
A second option for running the code on a Linux/UNIX system is, once located in the correct directory, type
make runinto the terminal.
For running NeuralNetworks.py from the command line in a Linux terminal:
-
cd into the directory where NeuralNetwork.py, "train.csv", "train" directory, "test" directory, and "test_idx.csv" are located (these should all be in the same directory with the exact names listed).
-
Once in the correct directory, simply type
python NeuralNetwork.pyinto the terminal. -
For Windows, the process is the same, however to run, you will need to enter
py NeuralNetwork.pyand follow the prompts outlined as above. -
A second option for running the code on a Linux/UNIX system is, once located in the correct directory, type
make runinto the terminal.
For running CNN.py from the command line in a Linux terminal:
-
cd into the directory where LogReg.py, "train.csv", "train" directory, "test" directory, and "test_idx.csv" are located (these should all be in the same directory with the exact names listed).
-
Once in the correct directory, simply type
python CNN.pyinto the terminal. -
For Windows, the process is the same, however to run, you will need to enter
py CNN.pyand follow the prompts outlined as above. -
A second option for running the code on a Linux/UNIX system is, once located in the correct directory, type
make runinto the terminal.
For running Representation.py from the command line in a Linux terminal: *You will be prompted to give paths the you mp3 files and train/test csvs. Verify these paths and that they have appropriate data.
-
cd into the directory where Representation.py, "train.csv", "train" directory, "test" directory, and "test_idx.csv" are located (these should all be in the same directory with the exact names listed).
-
Once in the correct directory, simply type
python LogReg.pyinto the terminal and respond to prompts. -
For Windows, the process is the same, however to run, you will need to enter
py Representation.pyand follow the prompts outlined as above. -
A second option for running the code on a Linux/UNIX system is, once located in the correct directory, type
make runinto the terminal.
python --version
3.8.3
pandas.__version__
1.0.5
scipy.__version__
1.8.0
matplotlib.__version__
3.5.1
Tensorflow.__version__
2.4.1
librosa.__version__
0.9.1