genreAI

This project leverages machine learning techniques to classify audio files into different music genres. It uses a neural network trained on extracted audio features such as MFCCs (Mel-Frequency Cepstral Coefficients).

Features

• Classifies audio files into genres (e.g., rock, pop, classical, jazz).
• Open-source and available for contributions.
• Extracts audio features for analysis.
• Provides a web interface for easy interaction (coming soon).

Dataset

The dataset used for training the model is the GTZAN dataset, which contains 1000 audio tracks categorized into 10 different genres. The dataset includes extracted audio features (MFCCs) from these audio files. Find out more about GTZAN here.

Model

The model is built using TensorFlow/Keras. It consists of a simple neural network that processes the MFCC features and outputs the predicted genre.

Installation

To run this project locally, follow these steps:

1. Clone the repository:

git clone https://github.com/ikeawesom/genreAI.git
cd genreAI

2. Install the required dependencies:

pip install -r requirements.txt

3. Download the GTZAN dataset:

• From this repo OR Kaggle
• Extract the dataset and place it in the ./datasets directory.

4. Train the model:

python train.py

Usage

1. Modify the MODEL and FILE_PATH variables within app.py to use your preferred model on a sample audio file of your choice. The sample model genreAI_model.keras and audio file test.wav has been provided within this repo.

MODEL = "genreAI_model.keras"
FILE_PATH = "test.wav"

2. Run app.py.

python app.py
>> Genre is rock.

Contributing

Contributions are welcome! Please follow these steps to contribute:

1. Fork the repository.
2. Create a new branch using git checkout -b feature/your-feature.
3. Commit your changes using git commit -m 'Add some feature'.
4. Push to the branch using git push origin feature/your-feature.
5. Open a pull request.

License

This project is licensed under the MIT License - see the LICENSE for details.