Holographic tags, microprinting and even radio beacons woven into fabric have been used by fashion labels for years to help establish the authenticity of their products. Apparel makers will spend US$6.15 billion ($8.3 billion) on anti-counterfeit technologies in 2017, according to London- based researcher Visiongain, but the anonymity of internet shopping and the growing popularity of second-hand dealers is making the war against fakes harder.
“Even 10 years ago, a woman going to buy a second-hand bag would know very well that Chanel, Gucci and Prada don’t sell on the street corner,” says Susan Scafidi, director of the Fashion Law Institute at Fordham University in New York. “But now, with so much legitimate and illegitimate commerce occurring online, it is very difficult for consumers to tell the difference.”
The issue was highlighted last year when the International Anti- Counterfeiting Coalition suspended the membership of China’s biggest online retailer, Alibaba Group Holding, amid criticism that it and other e-commerce marketplaces were not doing enough to cull fakes. Alibaba founder Jack Ma did not help matters when he said Chinese-made knockoffs today can offer better quality than the genuine articles.
Second-hand online stores such as RealReal and Vestiaire Collective use experts with years of experience to determine the authenticity of the goods they buy and sell. It is a painstaking process that is not absolutely foolproof, according to some online reviews from customers who complain they have been sold counterfeits.
Entrupy says its camera magnifies objects 260 times, so features invisible to the human eye become telltale signs: misshapen stamp marks, tiny gaps in leather grain and paint overruns.
The device, which looks like a bulky flashlight with a wireless connection, can be leased for an initial fee of US$299. Monthly plans start from US$99. So far, about 160 businesses including pawn shops, wholesalers and online retailers have signed up.
“Today, everything is done by humans,” Entrupy co-founder Vidyuth Srinivasan says. “For businesses that are growing, that’s not a scalable solution.”
Srinivasan and two New York University researchers, Ashlesh Sharma and Lakshminarayanan Subramanian, started Entrupy in 2012, a year that was a turning point for computer vision.
A breakthrough in algorithms at a science competition called ImageNet vastly improved the ability of machines to identify everyday objects in photographs by using massive data sets to find patterns. It was a watershed moment for deep learning technologies that also underpin self-driving cars and better speech recognition software.
With some help from Yann Le- Cun, Facebook’s director of artificial intelligence research and an angel investor in Entrupy, Srini vasan and his partners started with a hunch that computers could be trained to look at pictures of luxury goods and extract a kind of genome, an essence of, say, a Fendi or an Hermès handbag.
The problem was that deep learning requires tons of data they did not have: None of the founders had a closet full of designer handbags, fake or otherwise.
After some unfruitful spy missions to the women’s sections of department stores, they convinced several New York second- hand shops to give them access to their inventories. Getting the knockoffs was easier: One of the co-founders brought a suitcase- full back from a trip to China. Entrupy’s database now has tens of millions of photographs from about 30,000 different handbags and wallets. The software learns as clients upload new pictures.
Srinivasan says the company has no relationships with any of the fashion brands whose products they authenticate. LVMH Moët Hennessy Louis Vuitton Se and other makers of luxury goods prefer not to acknowledge that there is a second-hand market for their merchandise.
In July, Entrupy raised US$2.6 million from investors led by a venture between Tokyo-based Digital Garage and Daiwa Securities Group. The money will be used to design a faster and more portable camera and add more brands to Entrupy’s list, according to Srinivasan, who says the company is also looking at other uses for its software.
“The technology works pretty well on everything except for diamonds and porcelain, because those are refractive and we use optical analysis,” Srinivasan says. “We’ve already tested it on auto parts, phones, chargers, headphones, jackets, shoes, and even crude oil.” — Bloomberg LP
This article appeared in Issue 797 (Sept 18) of The Edge Singapore.
