How to extract place, action, audio, cast features using the pre-trained models provided

[spaul13]

Starting from a custom movie shot, how to extract the place, action, audio, cast feature tensors from it.