Eyealike, announced a new product today dubbed, Eyealike Copyright. This 'patent-pending' search platform aims to automate the process of copyright protection for online video content owners. According to the release, the service will automatically distinguish and detect copyrighted content within a video. This is quite a step forward as other sites (like Youtube, mentioned in the release below) currently rely on keyword tags, watermarks, and other identifiers rather than automating the detection of content within the video itself. The theory, of course, is that this service will make it more difficult for copyright violators to post and share videos online.
We presume this will rely on facial recognition (check out their>Celebrity Match demo – screenshot below). In fact, it makes sense as they clearly already have a collection of celebrity images, and can currently search a collection of pictures to match to these celebrity images, it follows that they would be able to detect actors and actresses within online videos. I would imagine that would cover a fairly large percentage of copyright materials that are currently being pulled on a constant basis from video sharing sites.. Remember, last week' s Tom Cruise scientology video? I am sure that eyealike could have detected just playing with the demo.
In any case, it seems like an interesting product. I imagine it is only a matter of time, however, before the larger engines and sharing sites begin to utilize more and more speech recognition, image, and facial recognition technologies. We even heard last month from Marissa Mayer, the Google project manager for their Goog-411 service. We were told that even though the product currently is a $$ loss, it is designed to collect, index, and analyze speech samples for later use in speech recognition for video search. Everyzing also has a wonderful way to do speech recognition currently. It would be interesting to find out if they are working on such a product as well….
Here is the screenshot as promised:
DEMO 08 – Eyealike, an innovator of , today announced Eyealike Copyright to help media content producers quickly eliminate massive amounts of online copy infringement that result in losses of millions of royalty dollars everyday. The new solution is unique in its ability to automatically analyze every aspect of video content simultaneously by image, motion, and face to deliver unprecedented speed, scale, and relevancy of recognizing and flagging any video content that is being served illegally online. The technology is specifically designed to help enforce copyright laws across User Generated Content (UGC) websites, while also giving both content producers and expediters the necessary tools to support the 'fair use' doctrine, which protects people who use copyrighted material for scholarship or review. "Content distributors face an impossible balancing act: protecting copyrights and empowering a consumer's right to legitimate content," said Chris Shipley, executive producer of DEMO 08. "Eyealike is removing that complexity with a new visual-based search approach that can protect copyrighted content without placing the heavy burdens experienced today by content producers and content distributors."
A formal demonstration of the new Eyealike Copyright solution will be at DEMO 08 on January 30 at 3:35 p.m. PT, and throughout the conference at pavilion station number 25.
Ending the Copyright Law Wars Online
One need not look any further than the $1B lawsuit Viacom has imposed upon YouTube/Google to realize that copyright infringement is at the top of nearly every media content producers mind. In addition, a report published by the Motion Picture Association of America in May 2006 found that the major U.S. movie studios lost $2.3 billion to Internet piracy in 2005, which doesn't even account for infringement taking place throughout industries like television, music videos, and sports.
While Google/YouTube and MySpace, among others, have been using various forms of copyright filtering technology and even human editors, these solutions in many cases require the utilization of audio, pre-tagged text, meta data, watermarks, and people to identify video clips or certain objects like faces within video clips. The challenge with relying on these approaches is that they can't scale to automatically distinguish between outright copyright infringement and legitimate use of short sections of copyrighted material, especially as infringed content is edited, cropped, and recompressed several times.
"In many cases the toughest part of 'flagging' copyright infringed content is the actual finding of the content itself, which is often hidden within UGC," said Imad Zoghlamoi, chief technology officer of Eyealike. "Eyealike Copyright can find the infringed content immediately to help producers and UGC sites take necessary actions to have that material taken down and explore appropriate legal actions."
Zoghlami continued, "Our goal is to help make copyrighted material 'scarce' to benefit producers like Viacom and offer 'insurance' to content expediters like YouTube/Google that need to more quickly determine 'fair use' content and distinguish between what's legal and what's not."
A Unique Visual-Based Search Approach
Eyealike Copyright is powered by the patent-pending Eyealike Visual Search Platform (VS), which allows for highly intelligent indexing and analysis to process hundreds of images and video clips per minute by still objects, object movement, and facial recognition. Following four years of proprietary research and development supplemented by unique technologies and expertise from the University of Washington, Eyealike's strength lies in its fast pruning technology, image recognition and multi-relevance weighting algorithm for exact and similarity matching. Eyealike has applied this robust approach to automatically identify any particular object in a video frame by frame, whether it's a chase scene, an actor, a backdrop, or a love scene.
For example, if Madonna released a new copyright protected music video through her new promotion firm Live Nation to MTV, individuals could immediately upload that video content to a site such as YouTube, thereby infringing upon the rights of the content owner. While YouTube might take immediate steps to take down the content, end users would have likely already hidden the content before the content could be tagged, therefore making it nearly impossible for a site such as YouTube to identify and take down every piece of non-licensed content. Eyealike Copyright could immediately identify this hidden content in seconds with 95 percent accuracy and with a near zero false positive rate.
Pricing and Availability
Available immediately, pricing is based on an annual licensing fee and varies depending on CPU and storage size, modules purchased, and frequency of utilization.