An NLP project to generate novel music
tensorflow
2.4.0
selenium
to scrape?scipy
to signal process if needed
- Implement a data input pipeline (dataset needs to be found for this - leaning towards dataset from Mutopia. Might have to scrape the site for the download links)
- Implement 1D causal convolutions (PixelCNNs) and hence the entire wavenet model
- Experiment with model sizes and hyperparams
- Vary architecture, search if others work better
- Extend to speech generation?