jbarrow/CommonForms

CommonForms数据集是一个大规模、多样化的表单字段检测数据集，包含超过55,000个文档，超过450,000页。该数据集通过过滤Common Crawl中的PDF文件来构建，这些PDF文件包含可填写的元素。数据集包含了多种语言和不同领域的内容，没有一个单一领域超过数据集的25%。它旨在促进表单字段检测技术的发展。CommonForms is a large-scale, diverse dataset for form field detection, containing over 55,000 documents and more than 450,000 pages. The dataset is constructed by filtering PDFs from Common Crawl that have fillable elements. It includes a mixture of languages and domains, with no single domain comprising more than 25% of the dataset. It is aimed at fostering the development of robust form field detectors.

搜集汇总

背景与挑战

背景概述

CommonForms是一个大规模、多语言的表单字段检测数据集，包含约55,000份文档和超过450,000页图像，用于目标检测任务以预测文本输入、选择按钮和签名等字段位置。该数据集通过筛选Common Crawl中的PDF构建，涵盖多种语言和领域，且无单一领域占比超过25%，旨在推动鲁棒表单字段检测器的发展。