TUMO workshop - Reading and computing music with AI

This Git contains main code and tasks for the TUMO workshop "Reading and computing music with AI", led by Chahan Vidal-Gorène (Calfa) and Baptiste Queuche (Calfa).

Week 1: Optical Music Recognition (OMR)

Goal: Analyzing and recognizing scanned music scores to generate audio music files.

Objectives:

• Understand how neural networks work and their application to music recognition;
• Discover the main tasks in computer vision and their application to music;
• Learn how to build an AI project, annotate documents and train/evaluate a model.

Technical objectives: learning conda, YOLO, labelstudio, oemer.

Data: A specific focus will be made on Armenian music scores (mainly from Komitas).

Full instructions: see week 1.

Week 2: Music Generation using AI (Generative Networks)

Goal:

• Generating audio music files from music scores
• Music Generation using GAN and style transfer

Objectives:

• Understand how generative networks work;
• Understand object classification;
• Training generative networks;
• Building webapp for demo purposes.

Technical objectives: learning streamlit, tensorflow, prismRNN.

Data: A specific focus will be made on Armenian music.

Full instructions: see week 2.

Final project

Download final webapp with models running.