When facial recognition software identifies a model as a… giraffe. An improbable situation that the clothing brand Cap_able has managed to do, an Italian start-up that manages to fool recognition systems with patterns and designs designed by artificial intelligence.
The Manifesto collection has been specially designed for this. Jerseys but also pants, t-shirts and dresses, all very colorful. Chaque piece contains a motif, called “patch contradictoire” qui trompe les systemes: soit les cameras n’identifient pas la personne, soit les cameras l’identifient as une girafe, un zèbre, un chien ou l’un des autres animaux intégrés dans the reason.
The idea came from Rachele Didero, who went through the MIT Media Lab. In New York, she read that tenants had objected to the landlord wanting to install a facial recognition system in his building. With an engineer friend, they decided to “merge fashion design and IT” to “protect their data”, she explained to the US media. CNN.
With the help of algorithms.
To find the process capable of fooling facial recognition, there are two possible options. They can create the pattern and then use an algorithm to adjust it. Or they define the colors, size, and shape they want the pattern to be, and then ask the algorithm to create it.
They then test the images on one of the most widely used algorithms in facial recognition called YOLO. Last step, the creation of the garment. To do this, they use a computerized weaving machine according to a now patented process. Your clothing manages 60 to 90% of the time to fool recognition systems, but the software could also improve, perhaps even faster.
For the most curious, prices start at 300 euros. Cap_able is not the first initiative of its kind. Other brands use pixelated prints or glasses doped with iodine. During the World Cup in Qatar, creative agency Virtue Worldwide designed a face paint capable of fooling the country’s facial recognition cameras.
Source: BFM TV
