Berlin Artist's Camouflage Shirt Blinds YOLO Surveillance Cams

A button-down shirt covered in flowery green-and-pink blobs made a real-time YOLO detector drop its 'PERSON' box the moment a reporter held it up, and reinstated the box when he lowered it. That is the demonstration Berlin-based artist Simon Weckert gave 404 Media's Jason Koebler, and it is the pitch for what Weckert calls digital camouflage.

Weckert built the pattern the slow way. He installed YOLO, the widely used open-source object-recognition algorithm, on his own camera rig and iterated on random patterns using gradient ascent, watching the model's confidence score for 'person' fall. 'When YOLO returns 100%, I know the pattern obviously isn't working,' he told 404 Media, describing how he rotates parts of the design, flips colors, and re-tests until the label vanishes.

The catch is that the demo is against a specific target. 'The problem is that we don't know what technology is behind these camera systems, so therefore I cannot claim 100 percent that the t-shirt works against this specific camera system, but I can say I know it works for the very public and famous YOLO algorithm,' Weckert said. Berlin police will not say what runs on the object-recognition cameras they recently deployed at the Kotbusser Tor subway stop, the city's first such police rollout.

The shirt is also a moving target against a moving target. 'Every time there's a new YOLO, a new pattern will come out. Just like in fashion, the summer winter collection, here it's the every YOLO edition,' Weckert said.

He frames the point as psychological more than technical. 'It's to give them the feeling that there are actually certain tools we can use against surveillance.' Four experts in our Who's Who directory shared Koebler's report.