SciPostLayout
收藏资源简介:
SciPostLayout数据集由OMRON SINIC X Corporation和早稻田大学共同创建,包含7,855个科学海报及其详细的手动布局注释,旨在支持布局分析和生成研究。数据集内容丰富,涵盖多种字体、大小和位置的元素,如表格和图表,适用于评估布局分析和生成系统。数据集的创建过程包括从F1000Research下载PDF格式的海报,转换为PNG格式并进行手动注释。SciPostLayout数据集主要应用于科学海报的自动生成和布局优化,旨在减轻作者的工作负担并提高读者对论文内容的理解。
The SciPostLayout dataset was co-created by OMRON SINIC X Corporation and Waseda University. It contains 7,855 scientific posters accompanied by detailed manual layout annotations, with the aim of supporting research on layout analysis and generation. The dataset has rich content, covering elements such as tables and figures with various fonts, sizes and placement positions, making it applicable for evaluating layout analysis and generation systems. The dataset was developed by downloading scientific posters in PDF format from F1000Research, converting them to PNG format, and conducting manual annotations. The SciPostLayout dataset is mainly used for automatic generation and layout optimization of scientific posters, with the objective of reducing the workload of authors and improving readers' understanding of the content of academic papers.
数据集概述
数据集名称
SciPostLayout
数据集描述
SciPostLayout 是一个用于科学海报布局分析和生成研究的数据集。该数据集包含以下内容:
- 训练集:6,855 张海报图像
- 开发集/测试集:各 500 张海报图像
- 开发集/测试集:100 篇论文与 100 张海报配对
文件结构
scipostlayoutposterpngtrain.zip:包含 6,855 张训练集海报图像dev:包含 500 张开发集海报图像test:包含 500 张测试集海报图像train.json:训练集的标注数据dev.json:开发集的标注数据test.json:测试集的标注数据load_dataset.py:加载数据集的示例脚本
pdf:海报PDF文件(可选)
paperpdf:论文PDF文件jpg:论文JPG文件mmd:使用Nougat解析的论文MMD文件
数据集大小
- 1K<n<10K
语言
- 英语
许可证
- CC BY 4.0

- 1SciPostLayout: A Dataset for Layout Analysis and Layout Generation of Scientific PostersOMRON SINIC X Corporation 和 早稻田大学 · 2024年



