VLR-CVC/ComicsPAP

Comics: Pick-A-Panel是一个用于漫画理解的数据库，它包含了五个子任务，分别是序列填充、角色一致性、视觉闭合、文本闭合和标题相关性。每个子任务都提供了漫画面板的上下文和选项，以及相应的特征，以便模型能够学习如何继续漫画故事。数据集分为训练集、验证集和测试集，适用于不同的模型训练和评估阶段。

Comics: Pick-A-Panel is a dataset for comic understanding, which includes five subtasks: Sequence Filling, Character Coherence, Visual Closure, Text Closure, and Caption Relevance. Each subtask provides a context and options of comic panels, along with relevant features, enabling models to learn how to continue the comic story. The dataset is divided into training, validation, and test sets, suitable for different stages of model training and evaluation.