Site icon Kaptur

10 questions to a founder: Kuznech

Image recognition, visual search and content classification  have been around for a while with various degree of success. With mobile shopping exploding, as well as myriads of photo/video based platforms, it is now  at the core of almost any online experience. While Google, Yahoo, Facebook and Microsoft throw millions in research for their own gain, independent companies offer powerful solutions to the rest of the market.  4 years old Russia-based  Kuznech is one of them. We spoke with co-founder and CEO Michael Pogrebnyak to learn more:

Michael Pogrebnyak, Co-founder, CEO at Kuznech
Michael Pogrebnyak, Co-founder, CEO at Kuznech

-­‐ Explain Kuznech. What does it solve?

“Kuznech” is a shortened version of a Russian word «kuznechik», which means grasshopper. The vision of grasshoppers is very different from that of human beings and many other creatures. Grasshoppers have 5 eyes in total: three small eyes on the top of the head and two large compound eyes. Each compound eye is made up of thousands of very tiny eyes called ommatidia. These miniature eyes take in small portions of light from the full image that a grasshopper is observing and build a big picture out of these “data”. We thought that a grasshopper could be a great symbol for a company working on image, video and object recognition.

-­‐ What kind of companies use your technology?

Mainly, these are companies that manage large databases of visual information (images and videos): dating sites, social networks, video and image hosting platforms. For all these types of companies, we provide content moderation services, like nudity detection (checking whether the UGC contains adult content or not), video and image search, avatar moderation. In this field, we work with Russian’s largest social networks Odnoklassniki and VKontakte (Mail.ru Group).

Then, it’s e-­‐commerce market, where we present our mobile recognition solution: “See – Snap – Purchase”. For example, a US company PartsTown (they sell genuine OEM replacement parts for restaurants and commercial kitchens) and Russian HSC (Helicopter Service Company) use our recognition technology in their mobile applications.

And the third use case is logo recognition in images and videos. Here we look towards sports sponsorship market and work with marketing managers of sports clubs.

-­‐ In a white paper you offer, you wrote that visual search market will reach $6B by 2014. Where is it now?

The global image recognition market is constantly rising. The reasons are that nowadays many people are switching to smartphones equipped by hi-­‐resolution mobile cameras. Also, the technology is constantly advancing which makes it possible to deploy image recognition in various verticals: e-­‐commerce, mobile commerce, logistics, healthcare, government sector and defense. Besides, image recognition is widely used in Augmented Reality.

According to MarketsandMarkets research, the market grows at a CAGR of 21.6% and is estimated to reach $25.65 billion by 2019. So, by 2015 visual search market may reach around $12 billion (in 2014 it was $9.65 billion).

-­‐ There are other visual search engines on the market. How does Kuznech differ from its competition?

There are quite lot companies of different sizes, especially in North America, which is a most technologically advanced region, that compete on the field of visual search. Some of them are truly brilliant in terms of technological breakthrough and marketing efforts. Our advantage is that we can solve a really wide range of visual search tasks and successfully act in different market verticals (mobile recognition, face detection, face recognition, content moderation, video search). This means that our core technology is almost universal so it can be built in pretty different applications.

-­‐ Within your 50+ clients, what interesting patterns of usage have you seen?

One of the most interesting use cases of our technology is Merchandising automation service. Keeping track of thousands of product permutations in brick-­‐ and-­‐mortar stores is an extremely resource-­‐consuming task. We created an automated software that compares planograms (diagrams that show where and how products should be placed in an offline store) with actual window displays, detects «bottlenecks» and creates a conformance report. The service helps minimize the participation of merchandising staff in the centralized control over goods displays in online stores.
And almost every week we receive one or two e-­‐mails from internet users where they ask Kuznech to help find some person in social networks. It’s always very nice to receive such messages, and we always reply them individually.

-­‐ Lately Kuznech has entered the logo recognition space. What is your aim?

Images and videos flitting about social networks and the Web itself are a gold mine for potential data. But 85% of the textual data accompanying the visual content does not support the actual brands that are showing up in an image/video. Since no one ever includes the brand name in the keywords, finding a logo in social networks can be a real hassle. We analyze the data e.g. in Twitter after some sports games and see how many brands may go missing in the digital ecosystem if marketers use only text search. Visual search tools help uncover and quantify the hidden visual conversation around brands.

-­‐ Kuznech offers 8 different product? Which one is seeing the most traction and why?

Our three most popular products are Mobile Recognition, Video Search and AdultContent Detector. Although the last one is not a usual topic to talk about to a wide audience, I’d like to make a few remarks on it.To prevent the uncontrolled uploading and sharing of adult content is one of the biggest challenges of most social networks and video sharing websites across the world. Detecting prohibited adult content turned out to be a real scientific research for Kuznech. We developed an absolutely new approach based on the company’s own software solutions (logotype detection, text recognition, and scene classification) with the help of Convolutional Neural Networks (CNN).

Kuznech technology can also help with copyright infringement
-­‐ Have you ever had the temptation of launching your own app and why haven’t you done it?

We are working on it right now.

-­‐ What would you like to see Kuznech offer that technology cannot yet deliver?

To speak globally, it would be great to have mobile recognition installed by default on mobile gadgets. Getting back to Mobile recognition:  research show that nearly 3/4 of younger shoppers (so-called “Millenials”) want image search functionality incorporated into online and mobile shopping to make their shopping experience be more visual and intuitive. They are a new visual generation for whom images inspire purchasing process. So looks like that the reality where visual search penetrates almost every aspect of life is inevitable. We’ll see.

Photo by sonica@2006

Author: Paul Melcher

Paul Melcher is a highly influential and visionary leader in visual tech, with 20+ years of experience in licensing, tech innovation, and entrepreneurship. He is the Managing Director of MelcherSystem and has held executive roles at Corbis, Stipple, and more. Melcher received a Digital Media Licensing Association Award and is a board member of Plus Coalition, Clippn, and Anthology, and has been named among the “100 most influential individuals in American photography”

Exit mobile version