How GPU-Accelerated Computing Is Putting A Crimp On Counterfeiting

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
Black Friday is upon us, kicking off the holiday shopping season. While millions are scouring shopping malls for the best deals, GPU-powered technology from a promising startup, Cypheme, could be used to scrutinize the items in our carts to ensure they’re the genuine article. Cypheme’s quest began when the mother of one of its four cofounders nearly died from taking counterfeit medicine. He and three friends — who had been working together on a different project in Silicon Valley — decided to start a crusade against counterfeit goods.

Counterfeiting is a worldwide problem, particularly affecting three major categories of products: drugs and medicines, luxury goods (including alcoholic beverages) and consumer electronics. Each can pose risks to consumer and worker safety, from dangerous side effects to toxic components. Yet the counterfeit industry generates hundreds of billions of dollars a year, according to the United Nations Office on Drugs and Crime.
 
Definitely an interesting idea. It makes me wonder how their, "AI" works. Is it a massive image based system looking for discrepancies or is there something else behind it?
 
It's pretty crazy how accurate some of these counterfeit items are. In many cases you wouldn't even know they were counterfeit without having an actual version sitting next to it.

A few years ago I saw a pair of Sony MDR-V6 headphones on ebay for a price that was too good to be true. These are very popular headphones, you see broadcasters and production crew on the news and at sports games wearing them all the time. I bought the one off ebay out of curiosity, and because I also had a known authentic brand new pair to compare them to.

http://www.head-fi.org/t/510110/sony-mdr-v6-counterfeit-comparison-thread-w-pics

DSC05574.JPG


Some of the differences seem semi-obvious since the authentic pair is sitting right next to it, but otherwise it would be pretty difficult to tell.
 
^ That's kind of what I'm wondering. If you're only able to notice it from certain angles how is an end user going to know that. Does the software suggest one default image? Then, the paper allows the servers to calculate dimensions and provide feedback to the end user. It is an incredibly interesting concept and I'm wondering how accurate it is?
 
You must log in or register to reply here.
Back
Top