It's not exactly news that Google (STOCK QUOTE: GOOG) dominates text search on the Internet. Nor is it surprising that Google is working to broaden its command of image and video search. But Google is going about the latter all wrong.
The problem with searching for images or video through Google is that you have to describe the image and moving pictures you're seeking with a text description. You then refine your search with more text detail. See the problem? Your visual senses, when searching, become completely reliant on your verbal capacity, along with the text and metadata associated with an image or video.
A video-search competitor, Blinkx, attempts to offer better results by searching a transcription of audio corresponding with the video. Again, your search results are based on auditory clues, not visual.
Enter Bryan Calkins, 47, and Dr. Leonid Kontsevich, 49, founders of CogniSign.
The two San Francisco-based entrepreneurs teamed up in 2003 and launched their image-search portal, xcavator.net, for stock images in 2007. With their launch, they turned search into a visual experience.
"When humans are processing an object or a scene for recognition, their visual attention is jumping from key feature to key feature," says Calkins. "We have mimicked that part of the human visual-processing system with our trajectory technology. And it is the same thing, evaluating those salient features and their spatial relationship to one another. Then accommodating changes in scale, changes in viewing perspective and changes in position within the image."
For example, if you are searching for a martini, you start your search by typing "martini." A screen of images then appears. From this point forward, all refinement can be made through visual cues. If you want only drinks in a blue hue, you don't have to type the word "blue" -- you can simply drag a color-picker to the hue and intensity of blue you desire. If some glasses have a lime wedge on them and you wish to refine your search to only include drinks with lime wedges, you can "trace" simple lines over the salient objects in the image you wish to keep. This is where CogniSign's spatial recognition and other patented technology shines.
Calkins says CogniSign is in the process of acquiring a Series A round of financing, and he does not wish to disclose sales or investment figures. However, Calkins did say that the company, now with 12 employees, has experienced quarter-to-quarter growth in excess of 25% and he said he expects CogniSign to be profitable by the end of 2009.
Xcavator.net is currently a gateway to six stock photo companies indexing 7 million images, and Calkins expects the portal to exceed 10 million images with an additional three stock photo companies in the first quarter of 2009. Serving stock photos, however, is just the beginning. What makes CogniSign's technology so powerful is its scalability and flexibility. While its search methods may be different from Google's, CogniSign's revenue model is very similar.
Calkins hopes that by mid-2009, his company will begin implementing its technology for search-based ad targeting. Its technology will be able to place more precise ads with images that don't have keywords or accurate descriptions. The company also plans to launch its own video-search tool, which is already receiving industry praise, by the end of 2009.
"If somebody's watching a video of golf, our technology can correctly identify that this is golfing, not pictures of a park," says Calkins. "We know that if it's the PGA, they don't need our technology for advertisement placement , because advertising is placed by humans for the PGA, but more automated solutions are needed for the longer tail. We're talking about more random, possibly even user-generated video about golfing."
This is where the rubber finally meets the road for CogniSign, especially considering the explosion of social media and user-generated images and videos.
"Advertising has faced a lot of challenges in user-generated content or social media content," says Calkins. "But everyone wants to solve that problem, and we think with our technology, adding a visual component to it and adding better targeting we can help solve that problem."
Although CogniSign is on the crest of big success now, it has taken a while. Calkins says when they started in 2003, they figured the demand would have been really high by 2006 or 2007, but it's only recently that market demand has caught up with their solution. CogniSign isn't the first business in their search space. But like Google, CongiSign could quickly dominate with superior technology. As for the company's ability to last as long as it has, Calkins credits running it with a low burn rate of capital.
Going forward, Calkins says, CogniSign will transfer its technology into the mobile environment. In short, it will use its image-recognition technology coupled with camera phones and location-based technology (GPS and satellite triangulation) to allow mobile-phone users to search visually by snapping pictures. For example, if a user takes a picture of a building in San Francisco, the image recognition could send that information to a third-party service and give a restaurant recommendation within that building by being able to recognize both the building and the location.
"Our grand vision is making the Web and mobile a more visual place," says Calkins.