CSC412 Winter 2020 Final Project
This project aims to generate polyphonic melodies with a linguistic approach, comparing performance between variants of hidden Markov models (HMM) and an Encoder-Decoder network that uses long-short term memory (LSTM) cells. The main objective of this project is to construct pleasant melodies that sound indistinguishable from human-composed ones. The project evaluates the models' performance by both quantitative and qualitative measures and discusses possible areas for exploration.
python run_lstm.py -i data_dirSee python file for additional options for input.
Check HMM notebook for details
HMM outputs are located in HMM folder, generated using furelise by Beethoven.
LSTM outputs are located in LSTM folder, generated using compositions from beeth folder