P.D.R. F.N.R.S. Towards a Genealogy of Visual Forms

Maria Giulia Dondero (National Fund for Scientific Research/ULiège)

Towards a Genealogy of Visual Forms: Semiotic and Computer-Assisted Approaches to Large Image Collections

4-year Research Project Financed by F.R.S.-FNRS (2022-2025)

Abstract

This research project focuses on the genealogy of visual forms, that is the filiation of images within the framework of large image collections (“Big Visual Data”). More specifically, the project will study the revival and migration of forms initially found in Renaissance and baroque paintings held in European museums. Aiming to go beyond the studies on visual similarities that are currently being conducted in digital art history and in the digital humanities, this research on the genealogy of forms seeks to produce a diachronic analysis of the long period which sustained the transmission of forms from the Renaissance to the present day. These genealogies can be studied in large collections of paintings and contemporary fashion photographs available in databases such as the Web Gallery of Art and the Fashion Photography Archives of the Bloomsbury Fashion Central. The forms that will be examined in their revival/deformation/renewal are those of the gestures of the human body. These gestures constitute forms that are difficult to segment using the instruments that have been used so far in the field of computer vision (image processing). Such instruments favor a segmentation of the image according to the recognizable and static objects it may feature. In order to analyze gestures and, more generally, to transform the current ways of segmenting images, the project intends to work on the concept of “force”, drawing inspiration from studies of painting by Gilles Deleuze and René Thom. This project is also in line with Aby Warburg’s theoretical-practical perspective on the migration of motifs, and with Henri Focillon’s work on the genealogy of forms. More generally, it aims to reconsider the relation between contemporary quantitative and qualitative methods supported by computational analysis.

Three steps 

My research project focuses on the relation between contemporary approaches in Image Processing and the genealogy of visual forms in art history as conceived of by Aby Warburg (Atlas Mnemosyne, 2012 [1924-1929]) and by Henri Focillon (The Life of Forms in Art, 1992 [1934]). In my view, the projects of Warburg and Focillon were left unfinished due to the difficulty of detecting patterns across works from museums and collections which were both dispersed geographically and disparate in terms of their media and of the periods they covered. The increased digitization of artworks, the availability of online databases, and the computer processing of large corpora of images now make such projects technically feasible. 

Firstly, in this project, I will consider the ongoing research projects on image similarities such as the Media Visualizationproject of the Cultural Analytics Lab led by Lev Manovich, and the Replica project launched by B. Seguin at the EPFL’sDigital Humanities Lab in Switzerland. 

Secondly, I will evaluate how the future of the digital analysis of images may benefit from the ideas underlying the ambitious projects of Warburg and Focillon. 

Thirdly, I will draw from my latest book, The Language of Images. The Forms and the Forces (Dondero 2020), to propose my conception of forms and forces in images to Computer Vision, with the objective of making continuous bodily gestures analyzable in large collections of paintings and photographs. The solutions offered by this project are based on a method of image segmentation that underlies a wider philosophical reflection. This method finds its roots in the French School of Semiotics, especially in the tensive semiotics developed by Jacques Fontanille and Claude Zilberberg (Fontanille and Zilberberg 1998), as well as in Gilles Deleuze’s work on pictorial diagrams (Deleuze 2003) and in René Thom’s idea of conflicting forces in painting (Thom 1983).

Three kinds of corpora

I intend to analyze a corpus consisting of representations of human gestures and movements. Each gesture and movement represented has to be treated as being constituted by an equilibrium (or disequilibrium) between bodily internal forces like the ones in the faces of Bacon’s portraits (retention, constriction, dissipation, attraction, etc.) and externally directed movements (towards the surrounding space) around the human figure, in the landscape surrounding it. It’s important to study the internal energies and the external movements of the body in relation to the landscape. In this sense, it’s difficult to make an a priori distinction between gestures and external movements because the external movements are the developments of the gestures, the irradiation of their directions. 

Why gestures and bodily movements have been chosen as the focus of the corpus? Because they are present in every genre (except in still lifes) and in every representational theme and style, but do not constitute a class of images in themselves. This is noticeable in databases (see for instance the classification by WikiArt): Gestures elude every classification of images. 

The images I consider relevant for the constitution of the corpora come in two kinds, as they belong to two different media and periods of time:

1. Fine art painting from the Renaissance onwards. This branch of cultural production has been chosen for at least two reasons: the first one is that my project aims at harvesting the legacy of Warburg’s analysis of bodily movements in which he acknowledged the Renaissance as an era of concentration of influences coming from opposing cultures in Antiquity (Dionysian and Apollonian); the second one is that Renaissance painting and sketches are the starting point for a complex and sophisticated manner of representing the body, especially the feminine one;

2. Contemporary fashion photography in the field of advertising, which I consider to be an iconographical heir to classical painting, especially in the ways that bodies, be they represented at rest or in movement, are found arranged in a variety of creative manners. The close relation between Renaissance painting and fashion photography, as far as gestures are concerned, had already been noted by Berger in the 1970s.

My hypothesis is that fine art painting in the Renaissance inaugurated a special focus on the representation of gestures and movements, which contemporary fashion photography has inherited. The latter is one of the most inventive domains of image production, being very close to artistic ambitions and compositions, and being the one domain which cites, imitates, and spoofs painting iconographies the most. 

Two examples of this path from the Renaissance to contemporary fashion photography can be found in the following schema where two works by Leonardo da Vinci have been used (Figure 1): 

Figure 1. Cédric Honba, Images Experience, 2020. 

My research program has been planned as follows: I will establish a small starting working corpus that is constituted by twenty paintings and their sketches when applicable (corpus 1) and a reference corpus (corpus 2) which comprises the big data archives such as The Web Gallery of Art and similar image databases which contain the fine art paintings and the contemporary fashion photographs. By working corpus, I mean the small corpus that it’s necessary to select as a starting point for the analysis of gestures in the Renaissance and baroque painting. It will be contrasted with the reference corpusin order to find the networks of images that have a legacy with the images of the working corpus in terms of remake/rework/resurgence. I will obtain a final analytical corpus (corpus 3) which consists of what will result from the computational analysis contrasting the working corpus with the reference one. In other words, the analytical corpus (a sort of virtual museum) will be constituted by the genealogies of visual forms that will be revealed through the contrastive operations applied to the working corpus with respect to the reference corpus.