Stay in Grid: Improving Video Captioning via Fully Grid-Level Representation | IEEE Journals & Magazine | IEEE Xplore