【摘要】 视频版权问题面临巨大挑战。基于视频检索数据集CC＿WEB＿VIDEO与视频部分拷贝检测数据集VCDB，结合Vision Transformer、注意力机制等技术，对复杂时间、空间视频的相似性进行研究，提出基于ViT的视频细粒度相似性评估模型，使用基于图的视频重复片段定位算法定位同源片段，并在实验数据集上验证了该模型的有效性。更多还原

【Abstract】 Absrtact:Aiming at the great challenge of video copyright issues, based on video retrieval data set CC＿WEB＿VIDEO and video partial copy detection data set VCDB, combined with Vision Transformer, attention mechanism and other technologies, the similarity of complex spatiotemporal videos is studied, and a fine-grained video similarity evaluation model based on ViT is proposed. A graph-based video repeat segment localization algorithm is used to locate homologous segments, and the effectiveness of the model is verified on experimental data sets.更多还原