README.md

AI Audio task

This project demonstrates a simple pipeline for extracting features from audio files.

Installation:

To setup this project in your own environment, the following libraries need to be installed with python 3.*.:

• Librosa
• Numpy
• pytest
• setuptools

Run the following commands in your project's root directory:

1. Setup a virtual environment and select as interpreter

   $ python -m venv
   

2. Activate the virtual environment

   $ .venv\Scripts\activate
   

3. run the below to install the libraries

   pip install -r requirements.txt in your shell
   

Instructions:

1. Run the following commands in the project's preprocessor directory to run the pipeline.

   • To run ETL pipeline that loads the audio data, extracts fetures and stores in another directory

     $ python bin/preprocessor/Preprocessor.py preprocessor melspectrograms /path/to/dataset /path/to/save/directory

2. To run the unnit test:

   • Locate the test_pipeline.py file in the preprocessor directory
   • Make sure to update the dataset path and the output path. below shows where and how:

         
         @pytest.fixture(scope='module')
         def pipeline():
             print('------------------setup-------------------')
        
             pipeline = Pipeline(dataset_path='ABSOLUTE PATH TO DATA SOURCE DIRECTORY GOES HERE',
                             output_path='ABSOLUTE PATH TO OUTPUT SOURCE DIRECTORY GOES HERE',
                             sample_rate=22050,
                             mel_bands=10,
                             n_mfcc=10,
                             hop_length=512)
     
     
     

   • Go to your terminal and make sure to cd into the preprocessor directory and run the below command:

         pytest -k test_pipeline.py 
     

3. To install the package locally, make sure you cd into the bin directory or locate the setup.py file. Then run the below in the terminal.

       $ pip install .
   

4. To use the installed package in the terminal or batch, example is as shown below

       # you can change the feature type to mfccs or melspectrograms
       $ preprocessor melspectrograms /path/to/dataset /path/to/save/directory
   

Files

• bin

  • preprocessor
    • init.py
    • command_line.py
    • Preprocessor.py # pipeline class for preprocessing
    • test_pipeline.py # unit test class
  • setup.py # package setup

• music_dataset

• output

• README.md

• requirements.txt

Licensing, Authors, Acknowledgements

Many thanks to Valerio Velardo for this task.