This repository contains the workshop presented at the PyData 2017 conference in Barcelona.
The goal of the workshop is to familiarise intermediate python users with deep learning for audio processing. In particular, we deal with the instrument detection in polyphonic audio files.
In the first part, we introduce key concepts in deep learning.
In the second part of the workshop we introduce the data pre-processing steps (Jupyter notebook 'DataPreprocessing.ipynb') including audio file reading, feature computation and batch generation.
In the third part of the workshop we introduce the TensorBoard visualisation via Keras Callbacks (Jupyter notebook 'TensorBoardDemo.ipynb') including weights, gradients and graphs visualisation, embeddings and hidden outputs computation and visualisation.
python 2.7
jupyter
numpy, scipy, sklearn, pandas, matplotlib, keras>=2.0.4, tensorflow>=1.0