Semantic-Driven Saliency-Context Separation for Video Captioning | IEEE Conference Publication | IEEE Xplore