27 January 2026 in Multimedia Tools and Applications
PZPepe Ballesteros Zapata
DCDario Negueruela del Castillo
Dario Negueruela del Castillo
This paper explores the intersection of artificial intelligence and art curation through a project developed for the 2023 Helsinki Art Biennial, "New Directions May Emerge." Utilizing the Helsinki Art Museum (HAM) collection, our work re-imagines the city of Helsinki through the lens of machine perception. We employ CLIP to represent works of the HAM collection and assign fictional public coordinates to indoor artworks based on cosine similarity to existing outdoor public art. For each of these new locations, we generate synthetic 360° art panoramas with Stable Diffusion. This process approximates real-world spatial composition, derived from urban panoramas’ depth map extracted through MiDas, as a structural guide via ControlNet, while the aesthetic and thematic content is informed by CLIP-Interrogator-generated textual descriptions of the original artworks. The result is a machine curation that merges the artwork with a re-imagined physical space, inviting to blur the distinctions between art, context, and the machine’s interpretation. The project is presented as an interactive, web-based installation, inviting users to navigate a re-imagined version of the city and its cultural heritage. This work addresses how contemporary machine learning can move beyond data management to become an active participant in curatorial practice, creating new forms of audience interaction and spatial storytelling.
5 June 2023 in Humanities and Social Sciences Communications
Claude Debussy’s personal style is typically characterised as a departure from earlier diatonic tonality, including a greater variety of pitch-class materials organised in fragmented yet coherent compositions. Exploiting the music-theoretical interpretability of Discrete Fourier Transforms over pitch-class distributions, we performed a corpus study over Debussy’s solo-piano works in order to investigate the diachronic development of such stylistic features across the composer’s lifespan. We propose quantitative heuristics for the prevalence of different pitch-class prototypes, the fragmentation of a piece across different prototypes, as well as some aspect of the overall coherence of a piece. We found strong evidence for a decrease of diatonicity in favour of octatonicity, as well as for an increase of fragmentation accompanied by non-decreasing coherence. These results contribute to the understanding of the historical development of extended-tonal harmony, while representing a fertile testing ground for the interaction of computational corpus-based methods with traditional music analytical approaches.