SynSlide

该数据集由两部分子集组成，分别为SynDet和SynRet，是通过特定流程生成的合成幻灯片数据集。SynDet子集用于文档布局分析，包含了16个元素类别的自动边界框标注；而SynRet子集则专注于基于文本的幻灯片检索，包含了主题连贯的幻灯片图像及两种类型的摘要。该数据集规模宏大，包含了2,200张高质量的幻灯片图像，旨在支持文档布局分析以及基于文本的幻灯片检索任务。This dataset consists of two subsets, SynDet and SynRet, which are synthetic slide datasets generated via a dedicated workflow. The SynDet subset is tailored for document layout analysis, featuring automated bounding box annotations for 16 element categories. The SynRet subset focuses on text-based slide retrieval, containing thematically coherent slide images and two types of summaries. Boasting a large scale with 2,200 high-quality slide images, this dataset aims to support both document layout analysis and text-based slide retrieval tasks.

搜集汇总

背景与挑战

背景概述

SynSlide是一个合成幻灯片数据集，包含SynDet和SynRet两个子集，分别用于文档布局分析（16个元素类别的自动边界框标注）和基于文本的幻灯片检索（主题连贯的幻灯片图像及两种摘要），总计2,200张高质量图像。