amrithagk/sakuga_preprocessed

该数据集包含多个配置（seg1到seg5），每个配置包含多个特征，如标识符、哈希标识符、URL链接、场景开始和结束时间、帧数、关键帧数、动画标签、用户标签、文本描述、美学评分、动态评分、评分、文本概率、宽度、高度、文件扩展名、帧率、分类信息（时间、地点、媒体、拍摄、构图、角色）、索引级别、视频ID、视频段号、字数等。数据集分为训练集、验证集和测试集，每个集的大小和样本数都有详细说明。

The dataset contains multiple configurations (seg1 to seg5), each with various features such as identifier, hash identifier, URL link, scene start and end time, frame number, key frame number, anime tags, user tags, text description, aesthetic score, dynamic score, rating, text probability, width, height, file extension, fps, taxonomy information (time, venue, media, filming, composition, character), index level, video ID, video segment number, word count, etc. The dataset is divided into training, validation, and test sets, with detailed information on the size and number of examples for each set.