Video Summarization Through Reinforcement Learning With a 3D Spatio-Temporal U-Net | IEEE Journals & Magazine | IEEE Xplore