Clothing designer tricks AI-powered face recognition

An Italian clothing startup unveiled a men’s and women’s line that can evade facial recognition technology by tricking the AI-powered algorithm into thinking those wearing the garments are animals.

Milan-based Cap_able recently rolled out its “Manifesto” collection of hoodies, pants, T-shirts, and dresses for those seeking to protect their privacy.

“In a world where data is the new oil, Cap_able addresses the issue of privacy, opening the discussion on the importance of protection from the misuse of biometric recognition cameras…,” the company writes on its web site.

“Our Manifesto Collection is the first design collection that offers knitted garments that shield facial recognition,” according to Cap_able.

The collection is tailored in a way that the clothing items “are not recognized as such by real-time recognition cameras.”

The clothing patterns “confuse the algorithm” by “making the camera detect dogs, zebras, giraffes, or small-knitted people inside the fabric, instead of the person wearing the clothes.”

Cap_able produced a video which offers a side-by-side comparison of people wearing its clothing and those who wear other brands.

“The cameras placed in public places do not recognize people wearing Cap_able as people, therefore they do not collect their biometric data,” the company said.

The privacy issue has gained traction in the Big Apple after Knicks owner James Dolan used facial recognition technology to block attorneys from law firms with active litigation against his company from entering venues he owns.

Last week, elected officials in New York warned Dolan-owned Madison Square Garden Entertainment that it risked government support by employing facial recognition software to ban individuals from venues such as Radio City Music Hall and the Garden.

Cap_able, founded by 2019 FIT graduate Rachele Didero, said that its technology is “based on the use” of so-called “adversarial patches” — or images that “make recognition systems ignore the other subjects in the scene and make them detect an item of choice.”

The clothing specifically seeks to evade “YOLO” facial recognition technology, which is considered the fastest real-time object detection system in existence, according to Cap_able.

“YOLO” — which stands for “you only look once” — is an open-source face detection system that uses neural networks to read objects with greater speed and accuracy than other technologies on the market.

Maintaining privacy will certainly come at a cost. Cap_able’s knit short sleeve crews will set the customer back some $310 while a long-sleeve hoodie retails for around $460.

“The Manifesto collection is an example of Cap_able’s modus operandi and broader project,” according to the company.

“It wants to educate the population on the importance of privacy and human rights by addressing the problem of misuse of facial recognition technology.”