Printed in plain sight

About the project

Prints are art historical uniques, precisely because they are not unique pieces, but copies. Digit-ized images are widely available, spread across collections all over the world. These digital copies can be collected and analyzed by computers. The analysis of ‘deep features’ – hidden, significant aspects of the image stack – makes it possible to reconstruct the surface of the original printing plate and to study the artist’s drawing with unseen detail.

Recent advances in artificial intelligence demonstrate the potential of neural networks, also for traditional print scholarship. Machine learning and computer vision enable batch comparisons between the many versions of prints. They assist in the quantification of stylistic evidence in questions of disputed attribution, and in the virtual reconstruction of the prints, which can be used for conservation purposes.

Aim of the project

By applying methods of unsupervised segmentation, feature extraction, texture mapping and pattern matching, automated classification of prints is feasible, even without metadata. Style identification, technique recognition, and iconographic linking are obvious applications. Our aim is to design a framework for the required data processing pipelines for this purpose. High-precision image alignment of digitized prints to allow comparisons remains an important prerequisite and the project’s main research challenge.

Preliminary experiments with woodcuts by Dürer have been successful. This project focusses on Rembrandt’s technically more challenging etchings and drypoints. As a proof-of-concept we plan to produce high-fidelity copperplates for Rembrandt’s top pieces, which can be used to make new impressions (replicas) that will also be in the interest of museum education.

Related events

2+3D Photography – Practice and Prophecies. Amsterdam, Rijksmuseum, 20-22 April 2022 (Due to Covid pandemic postponed from 2021).

Partners and sponsors

This Fellowship is made possible by the Migelien Gerritzen Fund/Rijksmuseum Fund, and is part of the Rijksmuseum Fellowship Programme.

Gerelateerde projecten

• Pushing the boundaries of image processing and visualization for cultural heritage: Deep neural network and photogrammetry as tools in image research.

• Drawing out Rembrandt (DoRe): Technical and art-historical research into the drawings by Rem-brandt, especially the pen drawings in the Rijksmuseum

• Colour printing ink: The application of non-destructive analytical techniques for the identification of colour printing ink

• Van Leyden Collection: The ‘Princely’ Print Collection of Pieter Cornelis van Leyden (1717-1788) and the founding of the National Print Room

Publications

• A. Lugmayr, M. Danelljan, L. Van Gool, R. Timofte, 'SRFlow: Learning the Super-Resolution Space with Normalizing Flow, ECCV 2020 Spotlight.' in: Preprint arXiv:2006 (2020) 14200v2.

• X. Shen, F. Darmon, A. Efros, M. Aubry, 'RANSAC-Flow: Generic two-stage image alignment. 16th European Conference on Computer Vision' in: Preprint arXiv:2004 (2020) 01526.

• G. Knaus, R. Stein, A. Kailus, LIDO-Handbuch für die Erfassung und Publikation von Metadaten zu kulturellen Objekten. Band 1: Graphik. Heidelberg (2019) pp. 152

• A. Stijnman, 'Dans l’atelier du maître: Les techniques de gravure de Rembrandt.' In: J. Rutgers, A. Stijnman (eds). Rembrandt graveur: La comédie humaine. Grenoble: Éditions Glénat, (2019) pp. 28–39.

• E. Hinterding, J. Rutgers, Rembrandt. The New Hollstein Dutch & Flemish Etchings, Engravings and Woodcuts 1450–1700. Ed. G. Luijten, 7 Vols, Ouderkerk aan den IJssel (2013).

• A. Stijnman, Engraving and Etching 1400–2000: A History of the Development of Manual Intaglio Printmaking Processes. London (2012) Archetype; Houten: Hes & De Graaf.

• L. van Tilborgh, T. Meedendorp, E. Hendriks, D.H. Johnson, C.R. Johnson Jr., R.G. Erdmann, 'Weave Matching and Dating of Van Gogh’s Paintings: An Interdisciplinary Approach.' In: The Burlington Magazine, vol. CLIV (2012) pp. 112-122.