Web application implementing and visualising state-of-the-art methods for melody extraction
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
algorithms
static
.gitignore
.gitmodules
README.md
app.py
examples.txt
wsgi.py

README.md

Melody Extraction

Web application implementing and visualising state-of-the-art methods for melody extraction

About

This is a semestral project that precedes my bachelor thesis on the same subject.

Installation

To install virtual environments (needed for running the algorithms), please run algorithms/environments/install-environments.sh.

The SourceFilterContoursMelody algorithm requires Essentia which needs to be installed to the respective virtualenv.

The application itself requires Python 3.6 and Flask. You can install it by running: pip install flask

Also shntool is required for uploaded wav file verification

Running

To run the server execute: python app.py

goals

  • frontend
    • create a graphical interface for visualising and interacting with selected algorithms
      • uploading audio
        • optional: add youtube as a possible source for extraction
      • setting parameters using a custom API
      • plotting extracted melody on piano-roll
        • plotting gold standard, if available
    • compare evaluation results between algorithms
  • backend
    • automated evaluation of selected algorithm according to MIREX evaluation procedure

technologies

  • front-end: javascript *

selected methods

Upload audio sanitizing, checking formats, … Evaluate extraction, including OrchSet incl. visualization MIDI visualization State of the art metody, API pro nové extrahovací metody Baseline extrakce Případně: Batch processing of data? Syntetická data?