A dataset of Dunhuang manuscript images for fragments conjunction

<p indent="0mm">The Dunhuang manuscripts are a precious cultural heritage created by various ethnic groups in China throughout history. Because of their antiquity, most of them manuscripts have suffered damage, making the conjunction work a key step of the Dunhuang studies. Manual conjunction is both challenging and time-consuming. With the advancements of computer technology in recent years, the automatic conjunction of fragments with computer assistance has become possible. However, there is a severe lack of high-quality image datasets due to the complex reservation conditions of the Dunhuang manuscripts, which are essential for the effective computer-assisted methods. In this dataset, we compiled a collection of high-quality fragment images based on published manuscript conjunction papers, and supplemented it with manually segmented images. The dataset comprises 95 groups of 366 images. Each group of data includes one complete image as a reference for conjunction, along with 2 to 7 fragments. The images primarily feature content in Chinese and some ancient Tibetan. The collection institutions include the National Library of China, the British Library, the National Library of France, and the Dunhuang Academy. The process of data collection and organization is standardized. Moreover, the fragment quantity of each group and the distribution of collection institutions are representative. Additionally, the quality of the fragment images is sufficient to support the training and validation of conjunction algorithms or models.