This repo contains materials relevant to the implementation of Deep Steps with a simplified Python code base displaying:
- Encoding an audio corpus using onset detection into binary many-hot encoded representation
- Encoding audio corpus to represent micro-timings in corpus as 48-PPQN offset values
- Defining deterministic 'stacked' Autoencoder neural network architecture
- Displaying and testing training of AE on pre-processed dataset
- Feeding new data forward through decoder to generate new rhythm data