Track features - Which models are used to extract track features?

[csipapicsa]

Hi,

Could you please provide information on the models Spotify uses to extract track features? I'm aware of similar models like Essentia (https://essentia.upf.edu/models.html). Are the models used by Spotify for track feature extraction publicly available, or is there a research paper that details them?

Thank you!

[csipapicsa]

Anybody?