Editing and Analysing Historical Astronomical Diagrams with Artificial Intelligence
Résumé
The EIDA project explores the historical use of astronomical diagrams across Asia, Africa, and Europe. We aim to develop automatic image analysis tools to analyze and edit these diagrams without human annotation, gaining a refined understanding of their role in shaping and transmitting astronomy. In this paper, we present a baseline method to detects lines and circles in historical diagrams, based on text removal, edge detection and RANSAC. We compare this strong baseline to a deep learning approach based on LETR. This work contributes to historical diagram vectorization, enabling novel methods of comparison and clustering, and offering fresh insights into the vast corpus of astronomical diagrams.
Licence : CC BY - Paternité