Digital Camouflage: Hawaiian Shirt That Tricks AI Surveillance
Berlin-based digital artist Simon Weckert has created a Hawaiian shirt designed to confound AI-powered computer vision systems used in surveillance cameras. The shirt, named Digital Camouflage, features saturated color transitions and overlapping shapes that disrupt the statistical patterns AI models use to identify humans. While it appears as an eccentric pattern to human eyes, AI detectors perceive it as nothing at all. Weckert developed the garment in response to the AI video surveillance planned for Berlin's Kottbusser Tor. He created the pattern using an adversarial loop with another AI, testing it with the open-source object-detecting model Yolo. The shirt exploits the fact that AI systems learn human appearance through statistical features like head and shoulder silhouettes and limb proportions, which can be broken by the shirt's design. Weckert explains that the pattern 'shouts over the quiet statistical whisper that says person,' making the wearer invisible to AI. This is not his first foray into tricking technology; he previously used 99 phones to fake a traffic jam on Google Maps. The shirt was featured in a Dezeen interview and photographed by Weckert.
Key facts
- Simon Weckert is a digital artist from Berlin, Germany.
- The shirt is called Digital Camouflage.
- It is designed to trick AI computer vision systems.
- The shirt was developed in response to AI surveillance planned for Berlin's Kottbusser Tor.
- Weckert used an adversarial loop with another AI to create the pattern.
- He tested the shirt using the open-source AI model Yolo.
- The pattern uses saturated color transitions and overlapping shapes to break the body's outline.
- Weckert previously used 99 phones to trick Google into thinking there was a traffic jam.
Entities
Artists
- Simon Weckert
Institutions
- Dezeen
- Yolo
- Cap_able
Locations
- Berlin
- Germany
- Kottbusser Tor