VRDU
收藏资源简介:
VRDU数据集是为视觉丰富文档理解而设计的新型基准,包含两个子数据集:Ad-buy Forms和Registration Forms,分别包含641和1915条记录。这些文档具有复杂的布局和多样化的模板,旨在模拟实际业务文档的复杂性。数据集通过高质量的OCR处理,并提供了详细的标记,包括实体的层次结构和重复性。VRDU数据集的应用领域包括自动化业务流程,如采购、银行和保险等,旨在解决从视觉丰富文档中提取结构化数据的挑战。
VRDU Dataset is a novel benchmark designed for visually rich document understanding, which comprises two subsets: Ad-buy Forms and Registration Forms, with 641 and 1915 records respectively. These documents feature complex layouts and diverse templates, aiming to simulate the complexity of real-world business documents. The dataset is processed with high-quality OCR and provides detailed annotations, including the hierarchical structures and repeatability of entities. The application scenarios of VRDU Dataset cover automated business processes such as procurement, banking, insurance and others, aiming to address the challenges of extracting structured data from visually rich documents.




