Self-Supervised Video Representation Learning with Meta-Contrastive Network.