Jointly-Learnt Networks for Future Action Anticipation via Self-Knowledge Distillation and Cycle Consistency | IEEE Journals & Magazine | IEEE Xplore