video-understanding

No frame left behind: Full Video Action Recognition

We cluster video frames end-to-end to use all frames in action recognition.