Can anyone explain how these feature (pickle) files are created?

[Coding511]

I have gone through the suggested GitHub pages from where the authors directly took the features. However, they have given the readymade feature files without scripts from scratch. Without understanding the feature extraction, no one can reuse the code, sir.

And the most important question is:

Iemocap has one video for multilabel audio files, how those video features are extracted for a dialogue.

Regards