Announcing AVA: A Finely Labeled Video Dataset for Human Action Understandingresearch.googleblog.com1 pointpgodzin9 years ago