ChinaOpen

ChinaOpen是由中国人民大学信息学院创建的多模态学习数据集，包含50k用于训练的自动标注视频和1k用于评估的手动标注视频。数据集来源于中国流行的视频分享网站Bilibili，旨在支持中文环境下的多模态学习和模型评估。数据集创建过程中，通过自动数据清洗和手动视频标注确保数据质量。ChinaOpen适用于评估模型在开放世界环境下的性能，特别是在中文视频内容上的应用。

ChinaOpen is a multimodal learning dataset developed by the School of Information, Renmin University of China. It comprises 50,000 automatically annotated videos for model training and 1,000 manually annotated videos for model evaluation. Sourced from Bilibili, a leading video-sharing platform in China, this dataset is designed to facilitate multimodal learning and model evaluation within Chinese-language contexts. During its construction, automatic data cleaning and manual video annotation are adopted to guarantee data quality. ChinaOpen is applicable for evaluating model performance in open-world scenarios, especially for applications targeting Chinese video content.