Musical Robot

A project that helps identify the genre of an mp3 music file and discover music of similar genres. See the demo at: https://www.youtube.com/watch?v=6ErHy6OuTg4

Repository Structure

• /docs/: Component specification, functional specification, and project presentations.
• /musical_robots/data/: Raw data that was used for training the ML model and data containing track and genre information.
• /musical_robots/tests/: Unit tests.
• /musical_robots/demo.py: Musical Robot User Interaction
• /examples/: Tutorial notebooks for SVM training and genre prediction.

Use

1. Download full repository.
2. Download the training dataset 'fma_small.zip' and the datasets 'fma_metadata.zip" from https://github.com/mdeff/fma into the data folder.
3. Set up the conda environment with "conda env create -f environment.yml"
4. Activate the conda environment with "conda activate data-exploration"
5. To use the service to identify an mp3 music file's genre and explore similar music:
   • Run the command "streamlit run musical_robots/demo.py" in terminal from the main repository.
   • The interaction will look as follows.

3. To replicate the ML model:
   • Follow the tutorial in 'examples/GenrePredictionTutorial.py'
4. To train your own ML model:
   • Follow the tutorial in "examples/TrainSVMTutorial.ipynb".
5. To run the unit tests:
   • Run python -m unittest discover -s musical_robots in the main directory.

Acknowledgements

Thanks to Prof. David Beck and Anant Mittal at the University of Washington for teaching the course and guiding us during the execution of this project.