Video search engines fail most of the time because they rely upon people to tag keywords of things that appear in the video. But VideoSurf is unveiling today a better way to search through videos. It is launching a beta search engine at TechCrunch50 in San Francisco.
The San Mateo, Calif.-based company is using computer vision technology to literally see into the content of videos and discern how to categorize the video.
It blends a mix of technologies so that it’s not only able to recognize who or what is in an image, but also to break down a video by frame and classify the frames by topic. As such, it takes on a host of competitors, from CastTV to Blinkx and Like.com, all off which have pushed forward on assorted variations of these features.
The VideoSurf search engine can recognize faces and group videos of celebrities and other people much more deftly than normal engines can. It displays a series of thumbnail frames from various scene transitions in the video to show the user exactly what is contained in the video. These video summaries tell the user at a glance whether he or she really wants to watch the video. You can also pick a particular thumbnail frame to navigate to the part of the video that is the most relevant to you.
Lior Delgo, chief executive and cofounder, said today’s search engines are too blind and can’t get past problems such as video spam, which are disguised to look like interesting videos. VideoSurf also has deals in place to aggregate video from sites such as YouTube, Comedy Central, ESPN and Hulu.com.
The software is smart enough to segment a video into different parts based on changes in the scenery. And it’s smart enough to recognize people and their associations. If you type in “stand-up comedy,” you get a video of George Carlin. The software grabs the most relevant video and ranks it first by looking at the air time and frequency of appearance of the subject of the search. It also returns videos with similar visuals.
With a better index for videos, VideoSurf.com hopes to become a portal for all web video. As it does so, it will compete with Invision TV, which debuted this week at the DEMOfall 08 show in San Diego. While VideoSurf’s technology looks impressive as far as it goes, pushing forward video search in ways that others haven’t, it doesn’t have the same social and personaliation features that Invision TV offers. Either way, for both of these players, they are taking on YouTube in their efforts to become a destination page for video, something that many other video search companies gave up a while ago because of YouTube’s momentum.
Delgo was previously the founder of FareChase, a travel search engine that Yahoo bought in 2004.
The secret sauce comes from VideoSurf’s chief scientist, Achi Brandt, who invented a mathematical formulas dubbed “multi-grid fast computation.” VideoSurf uses those algorithms to compute huge amounts of data in milliseconds. Eitan Sharon, chief technology officer, is an expert in computer vision.
The computation is done in the company’s servers, Delgo said. The company processes tens of billions of visual moments. The company has 22 employees and was founded in 2006. It raised $5.5 million in a round that included former Vice President Al Gore, CurrentTV chief executive Joel Hyatt, and others.
[Photo: Courtesy of NewTeeVee]
VentureBeat and marketing technology analyst David Raab are working on a new Marketing Automation usage and ROI study
. If you currently use a marketing automation system, help us out by answering the survey.
If you do, we'll share the resulting data with you.