Preprocess data

[vincentvanhees]

Here is a list of key steps in the pre-processing of the data (to be automated):

OpenFace

• Install in Linux: https://github.com/TadasBaltrusaitis/OpenFace/wiki/Unix-Installation
• Process mp4 input data: https://github.com/TadasBaltrusaitis/OpenFace/wiki/Command-line-arguments
• See ExploFace repository and other repositories in the emrecdem GitHub organisation for tools on post-process this.

Librosa

• Process with https://github.com/emrecdem/explibrosa
• See also the notebook tutorials inside this repository

For now I have started to put all key steps in one notebook which I will use as prototyping environment for database work: https://github.com/emrecdem/emrecdem-framework/blob/master/notebooks/Load%20time%20series%20and%20store%20in%20database.ipynb

To do:

• Turn it into a loop that runs over all files and creates a warning if no audio-video file pairs could be found.
• Down sample the time series data to ease initial exploration.
• Expand extracted feature overview.