U-DIADS-Bib

U-DIADS-Bib是一个专为古代手稿文档布局分析设计的数据集，由乌迪内大学的计算机视觉和人文研究专家共同开发。该数据集包含200张图像，每张图像都具有像素级别的精确标注，无重叠且无噪声。数据集内容涵盖了四种不同的手稿，每种手稿包含50页，这些手稿来自法国数字图书馆Gallica，包括6世纪至12世纪的拉丁和叙利亚语圣经。U-DIADS-Bib的创建过程涉及专家手动标注与机器学习模型的辅助，旨在提供高质量的标注数据。该数据集的应用领域主要集中在文档布局分析，特别是用于训练和评估机器学习模型，以解决古代手稿的自动分析问题。

U-DIADS-Bib is a dataset specifically designed for ancient manuscript document layout analysis, co-developed by computer vision and humanities research experts from the University of Udine. This dataset comprises 200 images, each with precise pixel-level annotations that are free of overlaps and noise. The dataset includes four distinct manuscripts, each containing 50 pages, sourced from Gallica, the French digital library, covering Latin and Syriac Bibles dating from the 6th to 12th centuries. The creation of U-DIADS-Bib involves expert manual annotation assisted by machine learning models, with the aim of providing high-quality annotated data. The main application areas of this dataset are centered on document layout analysis, particularly for training and evaluating machine learning models to address the automated analysis of ancient manuscripts.