3 ms·Announcing AVA: A Finely Labeled Video Dataset for Human Action Understanding1 points by pgodzin 9y ago