VIST-Edit
收藏资源简介:
该数据集名为VIST-Edit,包含了14,905个人类编辑过的视觉故事版本,这些版本源自2,981个由两种最先进模型——GLAC和AREL生成的机器视觉故事。每个机器生成的故事都对应着五个经过人类编辑的版本。该数据集按照80%的训练集、10%的验证集和10%的测试集的比例进行划分。这些故事是由GLAC和AREL模型生成的,数据集还包括了对这两个模型人类编辑的分析。规模上,从2,981个生成的故事中扩展出了14,905个编辑过的故事版本。该数据集的任务是视觉故事的后编辑。
The dataset named VIST-Edit contains 14,905 human-edited visual story versions derived from 2,981 machine-generated visual stories produced by two state-of-the-art models, GLAC and AREL. Each machine-generated story is paired with five human-edited versions. The dataset is split into training, validation, and test sets with a ratio of 80%, 10%, and 10% respectively. It also includes analyses of human edits targeting these two models. In terms of scale, the 2,981 generated stories are expanded into 14,905 edited story versions. The core task supported by this dataset is visual story post-editing.




