Multiple Sentences with Shots Dataset (MSSD)
收藏资源简介:
Multiple Sentences with Shots Dataset (MSSD) 是由快手科技和西安电子科技大学联合创建的一个大规模文本-视频对齐数据集,专门用于视频时刻蒙太奇任务。该数据集包含多句脚本和相应的视频片段标注,旨在通过文本描述精确匹配视频片段,并生成与文本描述一致的短视频。MSSD的创建过程包括从多个视频源中提取帧和镜头,并通过精细的标注确保文本与视频片段之间的高精度对齐。该数据集主要应用于短视频编辑领域,旨在解决基于文本描述自动生成高质量短视频的问题。
Multiple Sentences with Shots Dataset (MSSD) is a large-scale text-video alignment dataset jointly created by Kuaishou Technology and Xidian University, specifically tailored for the video moment montage task. This dataset contains multiple script sentences and corresponding video segment annotations, aiming to accurately match video segments with text descriptions and generate short videos consistent with the given text descriptions. The construction process of MSSD involves extracting frames and shots from multiple video sources, and ensuring high-precision alignment between text descriptions and video segments through meticulous annotations. This dataset is primarily applied in the field of short video editing, with the goal of addressing the challenge of automatically generating high-quality short videos based on text descriptions.




