Brit turns Comp Sci coursework into $100k glad-rags search biz

Vid A computer science graduate who used her final-year coursework on image processing to establish a startup has netted a $100k (£60k) prize.

Jenny Griffiths, who studied at Bristol University, became interested in computer vision during the second year of her master's degree - and in her fourth year she wrote software to recognise soft objects, such as clothes, and catalogue them.

Her image search system is now the core technology in her small fashion business that won Cisco's BIG startup award and prize money earlier this month.

Identifying fabrics and such things has been somewhat overlooked in the field of computer vision, which in terms of big commercial products has mainly focussed on medical and military applications. Griffiths instead angled her startup at the lucrative world of women's fashion.

Her resulting app - Snap Fashion - allows punters to photograph an item, or upload a picture of some apparel, and then search for similar ones across different retailers. This lets shoppers not only discover where a particular thing is sold, but also find the cheapest outlet and select variants of the things they more or less want. There's a link to buy the product, and Snap Fashion skims an affiliate link cut. The purchase can be done through the app or a common-or-garden web browser.

The software is geared at women's fashion, but there's an app for for men's stuff coming next. And potentially one for matching plumbing parts and such like after that. Here's a vid of the program in action:

"It's an idea we've all had, but I've backed it up with fairly complex algorithms that I invented in 2009 when I was at university," Griffiths said. "I tried to make a specific search engine that was particularly good at recognising colour and texture as well as on shape.

"I liked that with computer vision you could write some lines of code and get something back, instead of just a number. I guess I'm a visual person. I originally wanted to go into animation."

Obviously there are rival websites: Griffiths analysed eBay's search services and Google Product search, and researched competitors such as Tineye.com by reading the academic papers of its founders. Like Google image search, Tineye reveals where a submitted picture can also be found on the web.

But she's not scared about going head to head with Google: "It's complementary to what Google do. It's not a competitor. You'd be unwise to go into competition with Google, they're such a great company."

The tech at the heart of Snap Fashion is protected as a trade secret rather than being a patented design, which is unfortunate because it means Griffiths wouldn't actually explain to us how it works. Although she did tell The Reg in an interview that it was markedly different to the other techniques on the market: "There are so many different ways of doing computer vision that no one else is doing this."

But we can say that the software takes an image of, say, a jacket and defines the object as a blob of data values by identifying its shape and then colour and texture. The system then whips through a library of products to find items matching the definition and ranks them by similarity of shape and then of colour and texture.

Snap Fashion has a library of 250,000 items, and 120 retailers signed up on affiliate deals. Griffiths worked on establishing her startup in her spare time, outside of her full-time job, using money and expertise garnered from various small biz contests to build up the user interface, negotiate with outlets and assemble a team. Her startup also has Nokia's former UK veep onboard to guide it.

The Cisco BIG Awards: parking apps, video social networks and a bot that scan your Twitter to make sure you're being loyal to bosses

Snap Fashion took first prize at the Cisco BIG (British Innovation Gateway) Awards for startups at the finals held this month. The contest is part of Cisco's promise to leave a legacy after the London 2012 Games, and was built into its contract to provide networking gear to Olympic venues. The startup award judges received 200 entries, 30 reached the semi-finals and six made the finals.

The finalists ranged from software that streamlined workload in NHS hospitals - dubbed NerveCentre - to "community parking" app ParkAtMyHouse, and Shopitize, which seemed to rely on everyone taking photographs of their shopping receipts.

Second prize went to Digital Shadows - a small biz that attempts to find security vulnerabilities in corporate systems and produces an infographic-like report of its findings. THe company also studies employees' social networking profiles for data leaks, so it's hardly likely to be the most popular piece of software in the office.

And third place went to Six3, a video message service with the idea of replacing answer-phone messages with videos.

The Cisco BIG Awards will continue for the next five years in the same format. ®