页面结构分析数据集

页面结构分析任务以文档图像为输入，将文档图像分为几个大的区域，检测出其位置并给出其类型。本任务检测出的版面元素类型包括以下四种类型：版芯、页眉、页脚、边注（细分为版芯左侧和版芯右侧两种类型）。页面结构分析任务一般可以作为文档图像结构分析与内容识别任务的起点，可以有效地从文档图像中定位正文所在的位置，并决定页眉、页脚、边注等附属区域的相应位置，有助于后续任务聚焦于各自的重点区域。The page structure analysis task takes document images as input, partitions them into several large regions, detects their spatial positions, and classifies their types. The layout elements detected by this task include four categories: main content area, header, footer, and marginal notes, which are further subdivided into two subtypes: left marginal note and right marginal note relative to the main content area. This task generally serves as the starting point for document image structure analysis and content recognition tasks. It can effectively locate the position of the main body text in the document image, determine the corresponding positions of auxiliary regions such as headers, footers and marginal notes, and help subsequent tasks focus on their respective key regions.

搜集汇总

背景与挑战

背景概述

该数据集专注于文档图像的页面结构分析，核心任务包括检测文本块、页眉、页脚及旁注等布局区域，并扩展至列布局、浮动体位置、列表区域和公式块等多个子任务。这些分析任务旨在为文档内容识别与结构重建提供基础，支持后续处理模型如文本、表格和公式识别的精准应用。