AudioDup - Near-duplicate Detection of Audios

This repository presents my trivial approach for near-duplicate detection of audios, by generating acoustic fingerprints.

Setup Instructions

• We assume that you have access to a computer with MacOS. However, you should generally be fine with any Unix/Linux-based systems as well.
• Make sure you have installed Python 3.7 and the latest version of pipenv.
• Install MySQL connector using brew install mysql-connector-c.
  • Fix a potential bug by this.
• Install brew install portaudio && brew install ffmpeg.
• Install all dependencies with pipenv install.
• Setup a databset & user for the program:

CREATE DATABASE dejavu;
CREATE USER 'dejavu'@'localhost' IDENTIFIED BY 'dejavu';
GRANT ALL PRIVILEGES ON dejavu.* TO 'dejavu'@'localhost';

To Run the Program

• Collect fingerprints by pipenv shell python3 collect.py.
• Recognize sound from microphone by pipenv shell python3 recognize.py.

Testing

• We would use the FMA Dataset to perform testing. To avoid wasting too much time & disk space, you do not have to download the whole dataset.
• Put what you downloaded into the data folder.
• Run pipenv shell python3 collect.py to collect all fingerprints.
• Run pipenv shell python3 test.py to collect test results.

Licence

GNU General Public Licence 3.0