Mini-dataset for VL-Models fine-tuning (VL-Tune-dataset-mini)

A minimal dataset of 125 image-text pairs and 10 text queries for fine-tuning vision-language models on manuscript images. It is dedicated to the task of text-based image retrieval, and splited into "train" and "test" sets. The train set consists of 100 image-text pairs, while the test set consists of 25 image-text pairs. This dataset is constructed from the following sources:

- images from the DocExplore dataset of medieval manuscripts.

- images from two manuscripts from Al-Ḥarīrī, Maqāmāt, © Paris, Bibliothèque nationale de France. Département des manuscrits, namely MS arabe 3929 and MS arabe 5847.

- the descriptions in the text files are prepared by Martina Dinelli

The research for this work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy – EXC 2176 ‘Understanding Written Artefacts: Material, Interaction and Transmission in Manuscript Cultures', project no. 390893796. The research was conducted within the scope of the Centre for the Study of Manuscript Cultures (CSMC) at Universität Hamburg.