Each image was run through COCO-SSD, a pre-trained, off-the-shelf object-detection model built on the Common Objects in Context (COCO) dataset, a dataset that teaches machines to recognize everyday things (e.g., people, benches, cups, cars). But the usual machine vision logic is inverted: objects identified with high confidence are blurred, while those detected with low confidence—or not detected at all—remain clear. This draws attention to what machines struggle to perceive, exposing the blind spots and failures of artificial intelligence systems. By revealing how models are shaped by their training data–by these definitions of "common objects"–the project exposes the gaps that emerge from those definitions and datasets.

ABOUT THE CREATOR

Anna Zhang is an artist and creative technologist whose work explores how we relate to — and can reimagine — technology. Her work has been featured in Real Life, the New Media Caucus, BuzzFeed, and Forbes, and exhibited at venues including the National Museum of American History and Gray Area. She has been recognized by the Judson-Morrissey Excellence in New Media Award and Forbes 30 Under 30, is a member of NEW INC, and writes Negative Space.