RVL-CDIP MP, RVL-CDIP-N MP

RVL-CDIP MP和RVL-CDIP-N MP是由鲁汶大学的研究团队创建的两个大型多页文档分类数据集。这些数据集旨在解决现有数据集仅支持单页图像的问题，通过提供包含多个页面的文档来更真实地模拟实际应用场景。数据集包含多种文档类型，如信件、表格、电子邮件等，总计约400,000个样本。创建过程中，研究团队利用了OCR-IDL的元数据来匹配和检索原始文档。这些数据集的应用领域广泛，包括但不限于文档自动化处理、信息提取和分类，旨在提高文档处理的效率和准确性。RVL-CDIP MP and RVL-CDIP-N MP are two large-scale multi-page document classification datasets constructed by a research team from KU Leuven. These datasets are developed to address the limitation that existing document datasets only support single-page image inputs, enabling more realistic simulation of real-world application scenarios by providing multi-page document samples. The datasets cover a diverse set of document types, such as letters, tables, emails, and more, with a total of approximately 400,000 samples. During their development, the research team utilized metadata from OCR-IDL to match and retrieve the original source documents. These datasets have a wide range of application scenarios, including but not limited to document automation processing, information extraction and classification, with the goal of enhancing the efficiency and accuracy of document processing tasks.