Too confusing for AI person recognition?

Current image analysis software are sophisticated tools, supported by #artificialintelligence, learning and constantly improving. Nothing and no one escapes their attention, no matter if it's the pure recognition of people (that's ONE person) or their identification (that's THIS person). At least that's the impression you get over and over again. But how error-prone or easy to outsmart is this algorithm really?

Arguably, any technical development can be used for the benefit of humans or to their detriment, or more appropriately in this case, potentially to control them. The right and purpose of being permanently observed, recognized, stored and evaluated in public space without having any influence on all this personal information can certainly be argued. An Italian fashion start-up has chosen to give people their personal freedom as its mission. The tops and pants of the Manifesto design line are not only meant to be practical and chic, but also to let AI recognition go to waste by printing so-called "Adversarial Patch" patterns.

Thanks to the strikingly colorful knitted sweaters, pants, dresses, etc., automatisms identify people as animals or remove them as irrelevant "pattern sequences" directly from the analyzed video material. We don't know whether Cap_ablewill start a trend and pave the way for more personal freedom, but the software developers will certainly keep their eyes open and continue to work on recognition routines.