Urdu Music Genre Classification

For our deep learning course project we opted to work on a deep learning model that classifies Urdu music genres using convolution neural netwroks. The input was given in the form of MFCCs. In this experiment, a set number of random 30 second segments of a song were used to generate MFFCs. Further details are given in the report PDF.

Dataset

The data set consists of 4 balanced genres that have 250 songs each. MFFCs were calculated for each song and stored in data.json, the file that is used in the code.

Prerequisites

- Python 3
- Numpy
- Pandas
- Matplotlib
- Seaborn
- IPython Display 
- Scikit-learn
- Random
- Keras
- Tensorflow
- OS
- JSON
- Math
- Statistics
- Librosa

The code is provided in the form of a jupiter notebook that can easily be run on Google Colab or on a local environment.