Collaborative Development Workshop for Computational Research in Cultural Heritage, Art History, Visual Culture, and Digital Humanities
September 6–7, 2026, Malmö, Sweden
Supported by a grant from Riksbankens Jubileumsfond (Dnr F26-0091)
SCHEDULE
Sunday, 6 September 16:00 – 20:00
- 16:00-16:30 Welcome Reception
- 16:30-17:00 Amanda Wasielewski, “Introduction: Digital Art History and Computer Vision”
- 17:00-18:00 Stuart James / Nanne van Noord “Old and New Directions in Computer Vision for Art and Culture”
- 18:00-19:00 Discussion and mingle
Monday, 7 September 9:00 – 18:00
- 9:30-12:00 Art History and Heritage
- 9:30-10:00 Lia Costiner, “Recovering Portrait Identities and Tracing Artistic Collaboration with Computer Vision”
- 10:00-10:30 Frederik Mohammadi Norén, “A Reflection on Automatic Detection of Gender in Historical Photographs”
- Coffee 10:30-11:00
- 11:00-11:30 Ludovica Schaerf, “Art Beyond Semantics: Sheaf-Informed Contrastive Learning for Multi-Relational Representations”
- 11:30-12:00 Syrine Kalleli, “AIKON: a modular computer vision platform for historical corpora”
- Lunch 12:00-13.00
- 13:00-15:00 Digital Visual Studies
- 13:00-13:30 Eamonn Bell, “Why workflows for digital visual studies and why now? Experiences with Galaxy for the humanities”
- 13:30-14:00 Naja Grundtmann, “Style Transfer Across Domains: From Artistic Renderings to Medical Image Augmentation
- 14:00-14:30 Noa Garcia, “Generations of Style and Gender”
- Coffee 14:30-15:00
- 15:00-17:00 Visual Culture
- 15:00-15:30 Daniel Chavez Heras, “Pre-processing as Poetics: Modelling Film Editing as Space Puzzle”
- 15:30-16:00 Jan von Bonsdorff, “The Cobweb of Meaning: Visual Metaphor, Historical Advertising, and AI”
- Short break 16:00-16:15
- 16:15-16:45 Anna Näslund, “Attention Images”
- 16:45-17:15 Nausikaa El Mecky, “The Visual Tactics of Digital Ecofascism”
- Dinner 18:00
PARTICIPANTS AND ABSTRACTS:
Amanda Wasielewski
Amanda Wasielewski is Associate Professor of Digital Humanities and Docent of Art History in the Department of ALM at Uppsala University. She is the author of four monographs including Computational Formalism: Art History and Machine Learning (MIT, 2023)and Digital Photography After AI (MIT, 2026). She was awarded a grant from Google’s Artists + Machine Intelligence in 2023.
Stuart James
Stuart James is an Assistant Professor in Visual Computing at Durham University, Co-Director of the Durham Centre for Digital Humanities (DCDH), and Co-Director of the Institute of Medieval and Early Modern Studies (IMEMS). His interdisciplinary research brings together computer vision, artificial intelligence, cultural heritage, and the digital humanities, with a particular focus on visual and spatial reasoning across historical and cultural material. He has been a Co-I on major European projects including RePAIR, DCitizens, BoSS, and MEMEX, which explored themes ranging from cultural heritage and digital participation to social inclusion. Stuart regularly organises the Vision for Art and Culture (VISART) workshop and humanities-oriented tutorials that bring together researchers across computer vision, cultural heritage, and the humanities.
Nanne van Noord
Nanne van Noord is Associate Professor at the Multimedia Analytics lab of the University of Amsterdam. His research lies at the intersection of Multimodal AI and Visual Culture, with the aim of integrating equitable visual cultural understanding into AI models to bridge the gap between humanistic and algorithmic inquiry.
Lia Costiner
Recovering Portrait Identities and Tracing Artistic Collaboration with Computer Vision
Abstract:
This talk introduces two recently-launched projects that use computer vision to investigate distinct but related questions of analysis of variation in historical paintings. Faces of the Past uses deep-learning-based facial recognition to identify recurring sitters across a large corpus of portraits from the Low Countries, with the aim of recovering lost identities and enriching art-historical records. Semblance: A Computational Study of Artistic Collaboration in Renaissance Perugia applies computational methods to questions of compositional reuse and artistic collaboration in end of fifteenth-century Perugia. Presenting early results from both projects, the talk considers how computationally identified similarities can be assessed in relation to claims concerning identity, design transmission, and workshop collaboration.
Bio:
Lia Costiner is Assistant Professor of Art History at Utrecht University, specializing in late-medieval and early-modern visual culture, and digital methodologies. Rooted in the study of collections, her research explores how digital approaches can illuminate the creative and collaborative dynamics underlying artistic and technological developments of the past. She has held past positions at the University of Oxford, Villa I Tatti – Harvard’s Center for Renaissance Studies in Florence, and at EPFL, Switzerland.] She is currently principal investigator of two five-year projects supported by the Dutch Research Council and the Netherland eScience Center which she will discuss today.
Fredrik Mohammadi Norén
A reflection on automatic detection of gender in historical photographs
Abstract:
This presentation focuses on the challenges of using computer vision models to detect gender in historical photographs. Based on a failed attempt to map gendered spaces in 1930s Swedish photo archives, it instead explores the biases and limitations of models such as Grounding DINO, ViLT-VQA, and Llama. Drawing on distant viewing and critical data studies, the presentation shows how models trained on contemporary datasets often struggle to interpret historical imagery. Using a manually annotated validation set of 1,500 photographs from DigitaltMuseum, it demonstrates how automated gender classifications reproduce anachronistic assumptions, misrecognitions, and confirmation biases. Rather than treating these shortcomings as technical flaws alone, I argue that they can serve as productive objects of inquiry, revealing both the algorithmic logics of machine vision and the interpretive biases through which humans construct the visual past.
Short bio:
Fredrik Mohammadi Norén is an associate professor in media and communication at Malmö University in Sweden. His research interests concern media history, digital humanities, and propaganda history. He has co-edited several volumes, including Nordic media histories of propaganda and persuasion (Palgrave, 2022) and Media tactics in the long twentieth century (Routledge, 2024). His recent publications also includes the co-authored article ”On the Historical Gaze of Generative AI: Visions of Scandinavia in Stable Diffusion” (Scandinavian Journal of History, 2025).
Ludovica Schaerf
Art Beyond Semantics: Sheaf-Informed Contrastive Learning for Multi-Relational Representations
Abstract
Understanding a painting is never a single act. Art historians may analyze the same work through formal concepts, iconographic layers, and archival histories, to name a few, dimensions that are not interchangeable but each carry distinct relationships between the visual and the textual. Vision-Language Models like CLIP, which learn a single shared embedding space, collapse this richness into one homogeneous alignment, losing the multi-relational structure defining art-historical reasoning. We introduce CANVAS (Contrastive Art-aware Network for Vision-Language Alignment with Sheaves), a framework for learning relation-aware multimodal representations inspired by sheaf theory. Each entity is projected into multiple embeddings conditioned on the type of relation, and a novel contrastive loss encodes contextual structure during training, with no graph dependency at inference. We introduce new benchmarks of paintings and drawings for multi-relational art understanding: WikiArt+, derived from WikiArt and Wikipedia, and HertzianaDP. On multimodal retrieval and art understanding, CANVAS outperforms VLMs and graph-based adaptations, demonstrating that multi-relational alignment is not just theoretically motivated but practically essential.
Bio
Ludovica Schaerf is a PhD student in Digital Visual Studies between the Max Planck Society (MPG) and the University of Zurich (UZH). She holds a Bachelor’s in Liberal Arts and Sciences from Amsterdam University College and a Master of Science in Digital Humanities from the Swiss Federal Institute of Technology Lausanne (EPFL). Her interests lie in interdisciplinary research at the intersection of the Arts, Artificial Intelligence, and Philosophy. Her research focuses on framing the latent spaces of generative vision models from technical and media-theoretical perspectives. Ludovica complements her academic research with critical art practice focused on AI.
Syrine Kalleli
AIKON: a modular computer vision platform for historical corpora
Abstract:
Historical document analysis has progressed to a point where the main bottleneck for many historical applications is not algorithms, but relevant interfaces that can support historians’ workflow. While specialized tools exist for text processing and image search, we argue the community lacks a versatile collaborative platform enabling historians to analyze their own corpora from a particular perspective. As a step in this direction, we present aikon , a modular web-platform designed to empower historians with computer vision tools. aikon implements a complete workflow for historical document analysis, from corpus constitution to ai outputs validation and interpretation. It provides a comprehensive research environment combining source management tools with automated processing capabilities as well as multi-user validation and visualization interfaces. We showcase the potential of aikon by presenting modules enabling the investigation of graphical content transmission across large and diverse corpora, a problem for which few modern tools exist, and which is of key interest, for example in History of Science. Ségolène and Mathieu Aubrey led the work that Ségolène will present.
Bio:
Ségolène Albouy is a PhD student at the École des Ponts. In collaboration with historians, she is working on the dynamics of transmission in historical documents through the analysis of visual similarity.
Eamonn Bell
Why workflows for digital visual studies and why now? Experiences with Galaxy for the humanities
Abstract:
I describe recent developments by arts and humanities researchers to improve the reproducibility of digital humanities and social science workflows within the framework of the Galaxy Project (https://galaxyproject.org/). Galaxy allows researchers to define and execute computational workflows using free and open compute resource, and is sustainably funded by national and international research agencies. I illustrate some simple actionable workflows that are relevant to digital art history and digital visual studies and frankly describe the current state of this research software ecosystem and its associated community. These developments are put in the context of broader research policy initiatives advocating for and funding open science/open research practices. Finally, I describe some of the opportunities and challenges of engaging with these initiatives based on experiences gained in a number of digital research infrastructure projects with which I am involved.
Bio:
Eamonn Bell is Associate Professor in the Department of Computer Science at Durham University. His research interests fall under the broad umbrella of the digital humanities. These include: the application of mathematical and contemporary computational techniques to solve problems in the humanities; the history of the computational sciences and digital technology; identifying and lowering barriers to accessing digital research infrastructure (DRI) supporting computationally intensive research by non-specialists. Since 2019, his research has been funded by UK Research and Innovation (UKRI), the Irish Research Council, and a number of smaller institutional grants. He is most recently involved in the design and delivery of several DRI projects serving UK-based arts, humanities, and culture researchers, including DISKAH (https://www.diskah.org) and CCP-AHC (https://www.ccpahc.ac.uk).
Naja Grundtmann
Style Transfer Across Domains: From Artistic Renderings to Medical Image Augmentation
Abstract
Style transfer, a technique originally introduced for computational image stylisation, is now widely inspiring data augmentation approaches across fields from medical imaging to bioinformatics. This presentation explores what it is that makes this technique so readily transferable to other domains. Reflecting on the conceptualisation of style at the heart of style transfer, the presentation traces how style is reconfigured as an additive and functionally abstractable modality that can be modified independently from content. This, I argue, is what lends style transfer to being technically repurposed and operationalised in other domains, illustrated through a recent example from computational microscopy.
Bio
Naja Grundtmann is a researcher in the Department of Communication, University of Copenhagen. Her main area of research is located at the intersection of aesthetics, machine learning technology, and visual culture. She obtained her PhD Convolutional Aesthetics: A Cultural and Philosophical Analysis of the Perceptual Logic of Machine Learning Systems from the University of Copenhagen, Department of Arts and Cultural Studies in 2022.
Noa Garcia
Generations of Style and Gender
Abstract
Artistic style is more than a visual appearance. It emerges from specific socio-historical contexts and can encode the social hierarchies, identities, and gender norms of its time. Yet in computer vision research, style is often treated as a content-independent property, something that can be separated from semantics and defined through color, texture, or brushstroke. In this talk, I will explore what happens when we consider the social and historical dimensions of style in the generated image. I will begin by discussing the longstanding distinction between content and style in computational approaches to art. I will then turn to gender, asking how gendered representations are shaped not only by what images depict, but also by how they depict it. Building on top of this, the talk will explore the relationship between gender and artistic style across both historical artworks and contemporary generative images. Ultimately, I will show how generative models inherit and intensify gendered patterns from art history when prompted to generate images in specific artistic styles.
Bio
Noa Garcia is an Associate Professor at the Institute for Advanced Co-Creation Studies and D3 Center, The University of Osaka (Japan). She earned her Ph.D. in Computer Science from Aston University (UK), specializing in multimodal retrieval and instance-level recognition. She moved to Japan in 2018 as a postdoc, and has been conducting research at The University of Osaka since then. Her research sits at the intersection of computer vision, machine learning, fairness, and art. Her recent work includes investigating demographic bias in computer vision, analyzing visual datasets, and exploring how generative models can reinforce social stereotypes.
Daniel Chavez Heras
Pre-processing as Poetics: Modelling Film Editing as Space Puzzle
Abstract:
Computer vision approaches tend to struggle with long-range, structural temporality that characterises edited moving-image media. Standard pipelines encode video as sequences of frames or clips optimised for per-instance recognition, often discarding the relational structure between temporal units — the purposeful design of transitions, recurrence, and rhythm that is essential to editing films into coherent wholes.
In this talk I present a way to represent a film’s shot-to-shot transitions as a path through a state-space defined by shot-scale relations, adapting the “puzzle configuration graph” used in spatial reasoning to make editing rhythm computationally tractable as a signal and as a visible trace of how editing choices aggregate into styles and cinematic traditions. This works draws on FrameSense, a tool co-developed with King’s Digital Lab designed to make reproducible pre-processing pipelines for computational analysis of moving images at scale. Through this example, I will show how FrameSense data can be used to model film editing as a relational process, and how this can lead to better temporal representations in computational film analysis downstream.
Bio:
Daniel Chávez Heras is Senior Lecturer (≈Associate Professor) in Computational Humanities and AI Media in the Department of Digital Humanities at King’s College London. His work combines humanities frameworks in the history and theories of cinema, television, and photography, with applied technical practice in creative and scientific computing, including machine learning technologies. He is the author of Cinema and Machine Vision: Artificial Intelligence, Aesthetics and Spectatorship (2024, Edinburgh University Press), and is currently leading the project Intelligent Systems for Screen Archives (ISSA) funded by the British Film Institute.
Jan von Bonsdorff
The Cobweb of Meaning: Visual Metaphor, Historical Advertising, and AI
Abstract:
What if visual metaphor in advertising is not best understood as a neat source-target pair, but as a dense semantic web of products, bodies, texts, motifs, gazes, and absent archetypes? This paper introduces the poly-partite ‘cobweb’ model developed in ‘Mapping the Metaphor in Vintage Advertisements Using AI Tools’, a project based on Swedish mid-century advertisements. Emerging from the annotation of the Gabriella-brand visual corpus and guided by criteria such as decorum, metaphor fidelity, and historical tenacity, the model draws on mental spaces theory to capture how meaning is distributed across multiple interacting nodes. I argue that such a model is better suited to the complexity of visual and multimodal rhetoric than binary metaphor schemas. I also suggest that it points toward new applications in digital art history and visual culture studies, including machine learning workflows based on hypergraph representations of layered and partly implicit meaning.
Bio:
Jan von Bonsdorff is Professor of Art History at Uppsala University, where his work brings together art history, visual culture, and the critical study of digital methods. His research has ranged from medieval church art and nineteenth-century Scandinavian art to visual narration, scientific illustration, and intercultural exchange. In recent years, he has focused increasingly on visual metaphor, visual rhetoric, and the use of AI in the interpretation of images and cultural heritage. He currently leads the project Mapping the Metaphor in Vintage Advertisements Using AI Tools and is co-author, with Anna Foka, of AI and Image: Critical Perspectives on the Application of Technology on Art and Cultural Heritage (Cambridge, 2025).
Anna Näslund
Attention Images
Abstract:
In this talk I will present an ongoing sub-study from the research project ‘Selling Pictures. Pictorial Economies and Commoditization 1820s-2020s’, where we seek to historicize and contextualize the discourse on contemporary image production through generative AI.
By focusing images as mass produced products and their associated value – whether monetary, social and intellectual – it is possible to make analysis across different image techniques. This sub-study delineates and compares the discourses on the affordances of pre-photographic image techniques in the second half of the Nineteenth century (chromo-lithography, woodcut) with generative AI text-to-image-models in the early 2020s with a particular focus on attention. The product characteristics and values expressed in relation to pre-photographic techniques were primarily low price and large quantities of identical pictorial content. However the main function of these images in print media were to attract and hold the attention of the viewer much like the contemporary economy of visual attention.
Bio:
Anna Näslund is professor of Art History at Stockholm University and researcher at the Institute for Futures Studies in Stockholm. She has written extensively on photography, visual heritage and digitization. She is currently PI of the project ‘Selling Pictures. Pictorial Economies and Commoditization 1820-2020’ (Swedish Research Council/VR 2025-2028). Previous projects and publications include ‘The Politics of Metadata’ (VR 2019-2023), ‘Sharing the Visual Heritage’ (VR 2019-2023) and Critical Digital Art History. Interface and Data Politics in the Post-Digital Era (co-edited with Wasielewski, Intellect, 2024) and Travelling Images. Looking Across the Borderlands of Art, Media and Visual Culture (Manchester University Press, 2018).
Nausikaä El-Mecky
The Visual Tactics of Digital Ecofascism
Abstract:
How can harmful content appear pleasing to both the digital gaze and human audiences? How can a far-right movement present itself in such a way that it appears attractive and harmless while still swaying audiences towards its more violent ends? Focusing on the online presence of ecofascism in the early 2020s, the paper examines the self-contradictory nature of ecofascism’s online visual culture (including symbols, illustrations and certain emojis) and its ‘esoteric’ aesthetic, which is rooted in a visual and historical mish-mash that includes Nordic mythology and Nazi völkische art. The paper argues that the contradictory visual approach only made the movement and its dangerous content more effective in the online battle for attention.
Bio:
Nausikaä El-Mecky is associate professor in Art History and Visual Culture at Universitat Pompeu Fabra, Barcelona. She specialises in dangerous images across time—from the erasure of stone age statues to algorithmic suppression on social media. She is the author of the forthcoming monograph The Creation of Dangerous Images in Iconoclasm, Censorship, and Vandalism (Routledge, 2026) and co-editor (with Tomas Macsotay) of the volume Toppling Things as Memorial Contestation: Spectacle and Affect of Monument Removal (Brill, 2025). She was one of twenty female science champions selected worldwide by the Falling Walls Foundation (2025) and is the founder of Rebellious Teaching, a platform and community dedicated to unusual, rebellious approaches to education.
Funded with the generous support of Riksbankens Jubileumsfond (Dnr F26-0091)
