Multi-Annotation Attention Model for Video Summarization2023-06-010 PRE AI DOI:10.1109/cvprw59228.2023.00316原文链接原文求助分享收藏摘要 En