German artist Simon Weckert has created an unusual garment called “Digital Camouflage,” designed to interfere with computer-vision systems that automatically detect people. To a human observer, the shirt simply looks like a colorful abstract pattern. But its design was mathematically optimized to disrupt some of the visual information an object-detection algorithm uses to recognize a human figure.
Weckert developed the project amid the introduction of AI-assisted video surveillance at Berlin’s Kottbusser Tor. Modern object-detection systems don’t interpret an image exactly as humans do. Instead, they learn statistical combinations of shapes, edges, textures and colors associated with categories such as “person.” An adversarial pattern attempts to interfere with those signals.
The Digital Camouflage design was developed and tested using the open-source YOLO family of object-detection models. Demonstrations show the system placing detection boxes around other pedestrians while sometimes failing to register the wearer of the shirt as a person. Independent reporting by 404 Media also documented the effect during a demonstration.
There is an important limitation: Digital Camouflage is not a universal invisibility cloak. Weckert says performance can vary depending on the particular AI model, viewing angle, distance and lighting, and he does not claim that the shirt defeats any specific government surveillance system. Instead, he describes the project as a way to question how much trust should be placed in automated surveillance technology that the public often cannot independently examine.
