The project explores how painting can function as an operational design tool, using computer vision’s probabilistic readings to transform abstract paintstrokes into speculative architectural forms. Blurring the boundaries between author and algorithm, a custom-trained object detection system — trained using machine learning to recognise the author’s paint strokes — interprets these strokes, generating outcomes that are not as fixed objects but as a system of probabilistic translations.

Located on a neglected site along the Spree River near Berlin’s historic Eiswerk ice factory, the proposal reimagines the riverfront as an accessible bathing landscape rooted in the city’s cultural and industrial legacy.

The design process embraces the ambiguity, fluidity, and material unpredictability of painting, where architecture emerges as a temporal, evolving composition - guided by confidence scores from computer vision detections that modulate the balance between designer intent and machine uncertainty.

The result is a speculative landscape of possibilities and probabilities, shaped by the ambiguous, fluid qualities of painting.