Video Trimming Benchmark
收藏资源简介:
该数据集是为视频修剪任务专门创建的基准数据集,包含了从互联网上获取的原始用户视频,并标注了浪费和亮点标签。数据集的创建旨在通过视频字幕生成、片段过滤和故事编排三个阶段,生成具有连贯故事线的视频。数据集的应用领域主要集中在视频理解和视频编辑,旨在解决长视频中冗余内容过滤和关键片段提取的问题,从而生成更紧凑、更具吸引力的视频内容。
This dataset is a benchmark specifically constructed for the video trimming task. It comprises raw user videos collected from the Internet, annotated with waste and highlight labels. The dataset is developed to produce videos with coherent storylines via three stages: video caption generation, segment filtering, and story arrangement. Its primary application domains focus on video understanding and video editing, aiming to address the challenges of redundant content filtering and key segment extraction from long videos, thereby generating more compact and compelling video content.




