Rolling-Unrolling LSTMs for Action Anticipation from First-Person Video | IEEE Journals & Magazine | IEEE Xplore