News
The National Library of Luxembourg, always striving to offer the best services to its users, has launched a set of initiatives and experiments in 2022 aimed at optimising the visual exploration of our digital and digitised collections on the eluxemburgensia.lu site. To achieve this, we have relied on artificial intelligence (AI) technologies developed by experts in our IT and digital innovation division.
To make searching for illustrations, photos, portraits or caricatures easier, a proof of concept has been developed to classify images by genre and to identify those with strong similarities. Various properties of a set of 600,000 illustrations taken from digitised newspapers from 1844 to 2007 have been extracted using AI technologies. These properties have then been used to establish pre-existing similarities. A de-duplication process has also been implemented, to eliminate identical illustrations.
AI is also used to determine the content of these images. This development aims to simplify exploring illustrations from our digitised collections by adding keywords to the images.
Automatic learning models have been used to determine, for each illustration, its type (photo, drawing, etc.) and its general content (portrait, group, architecture, map, caricature, music, etc.). AI can also identify the objects present and the subjects describing the illustration.
With the aid of Google Vision models, 432 objects and 3,140 different subjects have been identified for 349,915 images.
While the development and refinement of these AI supported features are still ongoing, a first implementation on eluxemburgensia.lu available for the public is anticipated for the end of this year.
▪ This is an edited extract from our 2022 activity report, available for download on our publications page. Learn more
Last update