Abstraction
Heritage
Narrative
About
Where machines learn to perceive abstraction
The 3rd AI for Visual Arts workshop at CVPR 2026 explores how computer vision perceives, interprets, and models abstraction across artistic and cultural imagery—paintings, comics, sculptures, and installations that challenge assumptions of conventional models.
These stylized domains expose weaknesses in robustness, generalization, and interpretability, providing a principled framework for evaluating perception beyond realism. Can models "see" abstraction like humans, recognizing the dog-ness in a Picasso or inferring motion from comic lines?