you found my secret message :)

PatternPortrait

Description

PatternPortrait is an interactive art installation that merges artificial intelligence with robotic painting to explore authorship, style, and the creative agency of the artist. At the heart of the installation is a custom-developed graph neural network (GNN) that learns directly from a single pattern scribble and transforms it into a generative model capable of producing new line drawings in that specific visual language.
Unlike conventional AI image generators—which typically process rasterized, pixel-based data and are trained on massive and often intransparent image datasets—PatternPortrait takes a radically different approach. It works with vector-based path data, treating each drawing as a structured graph. This allows the GNN to directly model the relationships between strokes and paths, enabling a more fine-grained and interpretable understanding of artistic structure. The network is trained on only one individual example, preserving the distinct style and nuances of the artist’s hand. This minimalist training approach challenges dominant paradigms in machine learning and reflects a deliberate act of artistic self-empowerment: the machine does not create for the artist—it creates with them, as a collaborator grounded in the artist’s own visual vocabulary.
Visitors to the installation can experience this co-creative process first-hand. A quick photograph is taken of each participant, which is then stylized using background removal, edge detection, and finally processed through the trained GNN to apply the learned pattern aesthetic. The robot subsequently renders the stylized image as a physical line drawing on paper, allowing each visitor to witness a hybrid form of authorship—part human, part machine, fully informed by the artist’s creative signature.
PatternPortrait has been exhibited at several major venues, including the KI Festival Heilbronn (2023, 2024), the Zentrum für Kunst und Medien (ZKM) Karlsruhe, and the AI+X Summit at ETH Zürich. As an evolving research-based artwork, it is deeply rooted in bleeptrack’s PhD work in the fields of Machine Learning and Computational Creativity. The project bridges artistic intuition with technical innovation, questioning how AI systems can serve as extensions of an artist's style, rather than tools of anonymized automation.
By intentionally focusing on one-shot learning of single pattern drawings, PatternPortrait engages with current debates around dataset ethics, authorship, and the role of AI in creative domains. It highlights the possibilities of small-data approaches in artistic AI applications and opens up new conversations around what it means to train a model—not to generalize broadly, but to deepen the specific.

collection of portraits

Video Documentation

plotting in process

another close up

plotting software setup

On Display

On Exhibition:

Press + Mentions: