Anticipating Visual Representations from Unlabeled Video

Learning to See Slightly Ahead: Framing and Goals Motivation and high-level aim unlabeled video temporal structure anticipate actions anticipate objects At first glance, the promise here is simple: use the natural order of frames in raw footage so a ...