Our project will take sound as input, specifically music in MIDI file format. The output will be music generated from our model and played in the Malmo environment using note blocks. The goal of this project is to generate music that is inspired by the music that the model trained on. For example, if the model is trained on Bach’s music, the music generated will be similar to Bach’s. If the model trained on Bach’s and Mozart’s music, the music generated will be influenced by both composers.
Recurrent Neural Network (specifically LSTM)
The success of our project will focus primarily on the ability of the AI to generate music in Minecraft based on inputted music. As this is a creative project, there is no straight forward quantitative way to measure the success or failure of our model. Instead, we will listen to the music generated and assess the musicality of the music. We are exploring the possibility of looking at chords and sequences of notes to determine how close the music generated is to the genres trained on.
We should be able to discern progress towards the final goal by listening to the sounds being produced; if the sounds are completely incorrect, or if no progress is being made towards the ideal sounds, the algorithm might not be working as intended.
First appointment:
Date: 4/24/19 (Week 4, Wednesday)
Time: 4:15 pm - 4:30 pm
Second appointment:
TBD…