Action Recognition / Classification

Does fastai for pytorch (or other libraries in pytorch) have networks / architectures which are built /designed for learning action recognition in human motion? What I need to do is very similar to learning for example the ability to distinguish rolling eyes from eyes moving from side to side in face videos. Video length is between 5 to 10 seconds. Thank you.

I’m not sure about pytorch, but you may find some code of this kind of networks at

Thank you :wink: