Watermarks are useful tools for every kind of research and analysis concerning documents on paper. But it is necessary to know how to use them.
The “Watermarks for all” project aspires to make these resources accessible to a public perforce not specialised in this field.
The aim is to create a mobile application which allows to take a photo of a watermark, to upload it and obtain its data sheet, thanks to a system of pattern recognition, that asks directly for images with no need of words: people give an image and have an image back as an answer, together with some information about it. If the image doesn’t exist in the database, it becomes a new reference (crowdfunding).
The scientific approach consists in developing an algorithm based on the deep learning and the pattern matching methods (through convolutional neural networks), which needs multiple images of watermarks (identical or similar). In order to train the algorithm, a first database has been created from original photos.
The participants to the project:
- for the École nationale des chartes, Christine Bénévent, Oumayma Bounou, Ilaria Pastrolin, Olivier Poncet (person in charge), Marc Smith
- for the École de Ponts ParisTech, Mathieu Aubry
- for the INRIA – Institut National de Recherche en Informatique et en Automatique, Pierre-Guillaume Raverdy, Josef Sivic
- for the IRHT – Institut de Recherche et d’Histoire des Textes, François Bougard, Bruno Bon
- for the Archives nationales : Marie-Françoise Limon-Bonnet