Overview

We are glad to announce the Competition on FEw-Shot Text line segmentation of ancient handwritten documents (FEST) , which will be held in conjuction with the 19th International Conference on Document Analysis and Recognition ICDAR 2025.

Text line segmentation is a crucial component of Document Image Analysis, particularly for ancient handwritten documents. It directly influences subsequent tasks like optical character recognition, layout analysis, and information extraction. However, segmenting text lines in such documents poses unique challenges due to irregular handwriting, faded ink, and complex layouts with overlapping lines and non-linear text flow. Additionally, the lack of large annotated datasets makes fully supervised learning approaches impractical for historical documents.

This competition challenges participants to develop systems capable of segmenting text lines using only three annotated images per manuscript for training. The dataset features a diverse collection of ancient manuscripts with varying layouts, levels of degradation, and non-standard formatting, reflecting real-world complexities. By focusing on few-shot learning, this competition aims to encourage the development of methods that can be readily applied by humanities scholars without the need for extensive manual annotation, fostering broader adoption of automated document analysis tools.

Important Dates

• Registration Open: 10 December, 2024
• Test set release to participants: 7 April, 2025
• Deadline of Competition: 20 April, 2025
• Announcement of winner: 30 June, 2025

Info and Contact

• Silvia Zottin, silvia.zottin@uniud.it
• Axel De Nardin, axel.denardin@uniud.it

Partial financial support was received from Strategic Departmental Plan on Artificial Intelligence, DMIF, University of Udine.