MSc AI Project on generative deep networks and neural style transfer for audio
We prepared a short script called server_script.py in the learn_decay folder which hopefully contains a detailed enough description on how to train a model and generate a test sequence.
Here you can listen to our results or read our project report.
Oxford video lecture about RNN/LSTM Nando de Freitas
Keras for Sequence to Sequence Learning
An Empirical Exploration of Recurrent Network Architectures
Neat explanation of different LSTM variants with some good visualization
When and How to use TimeDistributedDense
Learning to generate text and audio
GRUV is a Python project for algorithmic music generation
fiala notes 1- Deep Learning and Sound ~ 01 Introduction
fiala notes 2- Deep Learning and Sound ~ 02: Generating Audio
Kaparthy's Stanford course CS231 on CNN
video - Simple(?) Step by Step
python - How to plot frequency spectrum?
Paper comparing different decay measures
Yaafe - package for audio feature extraction
Link to instrument files for first experiment Klick auf der Seite auf "Musical Instrument Samples/Post 2012/Woodwinds"
List of frequency ranges of many classical instruments
Stephan Mallat - Understanding Deep Convolutional Networks
JOINT TIME-FREQUENCY SCATTERING FOR AUDIO CLASSIFICATION
Junyoung Chung, Kyle Kastner, A Recurrent Latent Variable Model for Sequential Data
Variational Recurrent Auto-Encoder from Joost and Otto (UvA AI)
Justin Bayer & Christian Osendorfer, EARNING STOCHASTIC RECURRENT NETWORKS