ArtSleuth

ArtSleuth is a sophisticated computational framework designed to formalize the centuries-old practice of art connoisseurship through advanced machine learning. It meticulously examines the unique physical evidence painters leave on canvas, from brushstroke directionality and impasto relief to habitual gestures, leveraging self-supervised vision transformers to encode these subtle artistic signals. The framework introduces novel techniques such as style-guided cross-attention fusion for richer feature extraction, temporal style drift modelling to trace an artist's evolving hand, hierarchical workshop decomposition for probabilistic clustering of distinct contributors, and adversarial robustness testing to stress-test forgery detection against simulated historical techniques.

Características Principales

Casos de Uso