opendatalab/OmniDocBench

OmniDocBench是一个多样化的文档解析评估数据集，包含981个PDF页面，涵盖9种文档类型、4种布局类型和3种语言类型。数据集提供了丰富的注释，包括15种块级别注释和4种跨度级别注释，以及每个元素区域的识别结果和阅读顺序注释。此外，还包括页面和块级别的各种属性标签。

OmniDocBench is an evaluation dataset for diverse document parsing in real-world scenarios, containing 981 PDF pages across 9 document types, 4 layout types, and 3 language types. The dataset provides rich annotations including 15 block-level annotations and 4 span-level annotations, recognition results for each element region, and reading order annotations for document components. Additionally, it includes various attribute labels at page and block levels.