Palestrante
Descrição
J.M.W. Turner's atmospheric painting techniques resist categorical classification across two centuries, from Victorian imperial documentation to contemporary machine learning. As heritage objects whose institutional interpretation is now increasingly mediated by algorithmic systems, Turner's paintings sit at a critical juncture for heritage cataloguing practice. Reading Turner through Glissant's theory of opacity, this study asks when atmospheric dissolution constitutes resistance versus erasure. Blur-detection algorithms identified Turner's most atmospheric works from a corpus of 299 paintings. I then trained RF classifiers on photographic datasets to generate uncertainty maps that visualize where categorical systems fail. Structured visual coding quantified three atmospheric properties (dissolved edges, chromatic ambiguity, spatial flattening) while color distribution analysis measured computational resistance by comparing pixel-level uncertainty against photographic baselines. Turner's paintings push 21% more pixels into maximum uncertainty than atmospheric photographs. Three resistance strategies emerge: infrastructural withdrawal, material saturation, and luminous inversion. Whether implemented through Victorian documentary culture or neural networks, atmospheric dissolution disrupts the same taxonomic operations, operating as transhistorical computational resistance. The uncertainty maps materialize Glissant's opacity as quantifiable spatial phenomenon, showing how formal affordances operate transhistorically while carrying context-dependent political meaning. Turner's techniques function as erasure when applied to enslaved bodies, as refusal when protecting subjects from surveillance. Treating algorithmic failure as analytical method, the study makes categorical resistance spatially legible through uncertainty mapping. This has direct implications for heritage institutions integrating machine learning into collection management, where algorithmic cataloguing systems reproduce – and conceal – the same taxonomic failures. This methodology can be applied to other historical aesthetic strategies engaging contemporary algorithmic systems, connecting digital humanities to critical algorithm studies, postcolonial theory, and questions of surveillance and resistance.
Nota biográfica | Short Bio
Daniel Ungureanu is an Assistant Lecturer at the National University of Arts George Enescu Iași, Romania. He is a member and collaborator of various artistic research institutions (AAMG, BAN, ECREA, ICMA) and scientific journals (Leonardo, The European Journal of Humour Research; D:text Journal - Studies in Design Theory, History and Criticism), and has lectured at international conferences in Canada, Croatia, Portugal, South Africa, the Republic of Moldova, and the United States. He is the director and co-founder of the V-Cybercult International Conference, the author of Memes in Contemporary Visual Culture: Formative and Performative Aspects of Memetic Digital Content (2025), the co-author of How to Write a Text About Art (2022, with O. Nae) and Artificial Intelligence in Visual Arts Research (2025, with O. Nae), and the co-editor of Digital Politics of the Visual Global Age (2024; with C. Nae).
| Palavras-chave | Keywords | machine vision, algorithmic resistance, Turner, Glissant, heritage classification |
|---|