The genealogy of forms from the perspective of Deep Learning

International Symposium

Liège University

7th & 8th Novembre 2019

Salle de l’Horloge

Org. Maria Giulia Dondero (FNRS/ULiège) et Ralph Dekoninck (UCL)


The main contribution of the computational instruments used today in Digital Art History and, more generally, in the Digital Humanities devoted to heritage issues, is to reconnect with the project of a genealogy of forms that goes back to the mathematician and biologist D’Arcy Thompson (1917) and art historians Heinrich Wölfflin (1915) and Henri Focillon (1934) – and of course the notion of the migration of motifs in the work of Aby Warburg (1924-29). Yet the project of a genealogy of forms, despite various revisions and inclusions that are now taken into consideration, especially in the work of Didi-Huberman (2013), has remained unfinished because of the difficulty of detecting patterns in very large corpora obtained from museums and collections that are both dispersed and disparate (in terms of periods and media).

The increasing digitization of works of art, the availability of databases online and the computer processing of big corpora of images make this project technically feasible. Several studies in the United States and Europe involved in Deep Learning approaches using Convolutional Neural Networks (CNN) have demonstrated their effectiveness in visually recognizing series within large collections of visual documents by building signatures of objects of interest or of images as a whole (Feature Vector Signature). Deep Learning (Le Cun 2015) has thus demonstrated its unparalleled performance compared to visual wordmethods or more generally to methods based on extraction of local image characteristics. 

We can also mention the work of Lev Manovich (Manovich, Douglass, Zepel 2011) who analysed large collections of images through visualizations, thus making it possible to highlight the diachronic trajectories of the careers of several painters and to compare them (Manovich, 2015, 2017; Dondero 2017). Another example is the Replica projectof the EPFL Digital Humanities Lab, which aims to reconnect with Focillon’s project (di Lenardo, Seguin, Kaplan 2016) by using deep learning tools and, in particular, by expressing algebraic requests combining positive and negative examples to define the characteristics of the images sought. The objective is to reveal similar patterns and forms in groups of images that have not yet been linked by classical art history methods, and to update the mapping of cross influences. 

While most of this research aims to answer the questions raised by Focillon in the 1930s, some studies focus on the survival of Warburg patterns and forms (Hristova 2016). 

A new project, involving Belgium, France and Luxembourg, is being launched with the aim of taking up the theoretical and methodological aspects of the research programme on the genealogy of forms. The objective is to combine research in the field of advanced data-visualization technologies with semiotics research that has focused on the transmigration of forms (Basso Fossali 2013; Basso Fossali 2014; Dondero & Klinkenberg 2018-2019), and with the profound renewal of art history studies that embrace the formalism of their founders while reconnecting it to the study of meaning and opening it to a better understanding of the power of images. 

The aim of this two-day conference is to foster dialogue between art historians, semioticians and computer scientists on the classification of large image databases, based on comparison between the objectives and instruments of each disciplinary perspective.


Pierluigi Basso Fossali (2013), Il Trittico 1976 di Francis Bacon. Con note sulla semiotica della pittura, Pise, ETS.

Pierluigi Basso Fossali (2014), « Histoire des formes entre diachronie et archéologie », Actes du congrès 2013 de l’AFS (Association Française de Sémiotique) « Sémiotique et diachronie », Université de Liège, Belgique, 12-13-14 juin 2013. [En ligne] URL :, repris dans Vers une écologie sémiotique de la culture, Limoges, Lambert-Lucas, pp. 264-276.

George Didi-Huberman (2013), L’album de l’art à l’époque du « Musée imaginaire », Paris, Hazan. 

Isabella di Lenardo, Benoit Seguin, & Frédéric Kaplan (2016), « Visual Patterns Discovery in Large Databases of Paintings », Digital Humanities,DH 2016: Krakow, Poland, July 11-16, pp. 169-172.

Maria Giulia Dondero (2017), « The Semiotics of Design in Media Visualization: Mereology and Observation Strategies », Information Design Journal, 23/2, pp. 208-218. DOI:

Maria Giulia Dondero & Jean-Marie Klinkenberg (2018-2019), « Après Greimas. Des tâches pour la sémiotique visuelle », La part de l’œil, n° 32, Dossier « Greimas et la sémiotique de l’image » (Dondero & Klinkenberg dirs), pp. 230-235. 

Henri Focillon (1934), « Vie des formes », Vie des formessuivi de Éloge de la main, Paris, Presses Universitaires de France, 1943.

Stefka Hristova (2016), « Images as Data: Cultural Analytics and Aby Warburg’s Mnemosyne », International Journal for Digital Art History – DAH, 2.

Yann Le Cun, Yoshua Bengio & Geoffrey Hinton (2015), “Deep learning”, Nature, vol. 521, pp. 436–444.

Lev Manovich, Jeremy Douglass & Tara Zepel (2011), “How to Compare One Million Images?”, inB. David (dir.), Understanding Digital Humanities, London, Palgrave Macmillan, pp. 249-278.

Lev Manovich (2015), « Data Science and Digital Art History »,International Journal for Digital Art History, 1, 13–35. 

Lev Manovich (2017), « The Science of Culture? Social Computing, Digital Humanities and Cultural Analytics », The Datafied Society. Studying Culture through Data, Schäfer & van Es (Eds), AUP.

Aby Warburg (1924-29), L’Atlas mnémosyne, Paris, Éditions Atelier de l’écarquillé, 2012.      

Heinrich Wölfflin (1915), Principes fondamentaux de l’histoire de l’art, Éditions Parenthèses 2017. 

D’Arcy Thompson (1917), Forme et croissance, trad. de D. Teyssié, Paris, Seuil, 2009.

Organizing Committee

Pierluigi Basso Fossali (Lyon 2/ ICAR) 

Ralph Dekoninck  (UCL)

Maria Giulia Dondero (FNRS/ULiège)

Cédric Honba Honba (ULiège)

Scientific Committee

Pierluigi Basso Fossali (Lyon 2/ ICAR) 

Ralph Dekoninck  (UCL)

Maria Giulia Dondero (FNRS/ULiège)

Cédric Honba Honba (ULiège)

Pierre Leclercq (ULiège/LUCID)

Serge Miguet (Lyon 2/LIRIS)

Laisser un commentaire

Votre adresse de messagerie ne sera pas publiée. Les champs obligatoires sont indiqués avec *

Ce site utilise Akismet pour réduire les indésirables. En savoir plus sur comment les données de vos commentaires sont utilisées.