ImageRecognition
Latest
Yahoo buys image recognition firm IQ Engines to reorganize Flickr
Flickr is a killer resource for photographers of all stripes, but navigating through its massive photo catalog is far from elegant. Hopefully, Yahoo's purchase of IQ Engines can change that. According to the image recognition startup's website, IQ will be applying its skills to improving photo organization and search functionality to the online photo repository. Maybe its first order of business will be arranging all of Yahoo's new logos.
Googlerola buys Viewdle, ups Android's augmented reality and face recognition game
From existing tech like Face Unlock and Google Goggles to patent filings and Project Glass, it's clear that Google sees augmented reality and image recognition playing a big part in our computing future. It makes sense, then, that Big G subsidiary Motorola has bought Viewdle -- a Silicon Valley company that builds face, object, and gesture recognition technology for mobile devices. We don't know how much MMI paid for Viewdle, but we do know, thanks to a statement obtained by the good folks at TechCrunch, that the two firms "have been collaborating for some time." So, hopefully Android will reap the benefits (and fix those Face Unlock flaws) in the not-so-distant future.
NEC's Gaziru takes image recognition to the cloud, looks a lot like Goggles (video)
While Google's remained relatively quiet on the Goggles front, NEC's picking up where that image recognition left off with its own product, dubbed Gaziru. Showcased at Wireless Japan 2012, the company's angling its service, which aims to leverage both hardware- and cloud-based processing for smartphone queries, towards enterprise and consumer markets, highlighting its usefulness across a range of services from marketing to search. Much like the aforementioned Mountain View version, users would need only to snap a picture of an object with their phones to receive relevant search data, access product manuals or, in one scenario, car and real estate listings. Given its planned commercial launch this June, it won't be long before we'll get a chance to test this software en vivo. For now, content yourself with a translated video tour after the break.
Tobii looks to keep you alert by detecting when you're tweet-driving (video)
Don't think you'll get caught checking the Duke score while cruisin' down the highway? Soon, you may have to think twice before hitting the scoreboard. Tobii has unveiled its new technology that detects drowsiness and distraction in on-board driver safety systems. The platform is based on the company's advanced eye tracking tech to bolster automobile safety on the highways and byways. The system detects eyes of all shapes, sizes, and colors -- without calibration -- even if the driver is wearing glasses or a pair of Ray Ban shades. A constant stream of data communicates to the watchman the driver's condition, regardless of changes in environment or if the person behind the wheel takes a quick peek out the window. Tobii isn't looking to stop here either, as it says eye control of in-cabin infotainment systems is within reach. Perhaps this time next year, we'll be able to browse that Spotify collection with a series of blinks -- one can only hope.
Google Goggles Android update makes your vacation photos slightly more interesting
Google Googles' powers of perception have become a little more perspicacious, thanks to a new update for Android users. With version 1.6, tourists can use their smartphones to take a picture of a given area, while Google's visual search app works in the background to identify any notable landmarks, paintings or other objects. If it picks up on anything of interest, it'll automatically notify the user, instantly endowing him or her with gooey chunks of knowledge. It seems like a pretty user-friendly refresh, though things will really get interesting if faces ever get involved. Hit up the source link below to download the update for yourself.
StripeSpotter turns wild zebras into trackable barcodes
We've heard plenty of stories over the past few years about tagging animals with RFID chips, but we've never been particularly keen on the idea. Well, now a team of researchers has come up with a much less invasive way of tracking individual animals -- specifically zebras -- by essentially using their stripes as barcodes. StripeSpotter, as it's known, takes an isolated portion of a photograph of a zebra and slices it into a series of horizontal bands. Each pixel in the selection is then fully converted into black or white, and the bands are in turn encoded into StripeStrings, which eventually make up a StripeCode that resembles a barcode. All this information is stored in a database that allows researchers to directly identify particular animals without ever having to get too close. StripeCode may be a zebra-centric application for now, but its developers see it making a mark across the food chain with the inclusion of other distinctly patterned beasts, like tigers and giraffes. Animal tracking hobbyists can get their own free copy of the application by clicking on the source link below.
Junaio's augmented reality app for iPhone and Android can add cartoons to your otherwise-boring existence
As an augmented reality navigator that you want to use for finding pizza joints and friends on Foursquare and Twitter, you've got more polished options than Junaio's new 2.0 release -- Layar and Yelp, just to name a couple -- but Junaio has at least one interesting differentiator with a feature it calls "Glue." Basically, independent content providers can develop their own Junaio "channels," which function in the same way as a Layar layer -- it's a particular set of points of interest that'll be displayed in the current view. With Glue, though, Junaio isn't just using your positional information as a point of reference, it can also scan the image for specific objects that developers have programmed their channel to look for, and when such an object is found, crazy things can happen. In Junaio's demo, pointing the camera at a superhero-type cartoon dude causes a 3D representation of him to be rendered on the phone's screen; when you tap on him, he'll point his gun. It's an interesting concept that could eventually have some commercial implications, but in the short term, the company just needs to improve the app -- it crashed several times for us, and its support for changing orientation is a little wonky. If you want to check it out, it's a free download for both the iPhone and Android; while you wait for the install, check out our quick video demo after the break.
Harvard and MIT researchers working to simulate the visual cortex to give computers true sight
It sounds like a daunting task, but some researchers at Harvard and MIT have banded together to basically "reverse engineer" the human brain's ability to process visual data into usable information. However, instead of testing one processing model at a time, they're using a screening technique borrowed from molecular biology to test a range of thousands of models up against particular object recognition tasks. To get the computational juice to accomplish this feat, they've been relying heavily on GPUs, saying the off-the-shelf parallel computing setup they've got gives them hundred-fold speed improvements over conventional methods. So far they claim their results are besting "state-of-the-art computer vision systems" (which, if iPhoto's skills are any indication, wouldn't take much), and they hope to not only improve tasks such as face recognition, object recognition and gesture tracking, but also to apply their knowledge back into a better understanding of the brain's mysterious machinations. A delicious cycle! There's a video overview of their approach after the break. [Thanks, David]
Fujitsu announces 3D image recognition chip for robots
Fujitsu looks to be doing its part to keep robots well aware of their surroundings, today announcing a new 3D image recognition chip that promises an array of benefits, although it sadly doesn't have a robot to go along with it. Still, the company doesn't seem to think it'll have much trouble finding suitable bots for it, with its small size and low power consumption making it ideal for small, household robots in addition to those of the larger, more terrifying variety. Apparently, the entire system works in real time, and is said to be five times more efficient in recognizing images than similar system. While the chip has yet to find its way into an actual robot, that'll apparently happen before too long, with Fujitsu promising to put that final touch on the experiment before year's end.[Via Impress]
Penguinbot makes the perfect driving companion
It's always more fun driving with a buddy than cruising around alone, but since we lack what are referred to as "social skills," there aren't that many folks willing to ride shotgun in our car or hop on the handlebars of our Segway. Enter the robotic driving companion from Pioneer, a little penguin-shaped bot being shown off at CEATEC that perches on your dashboard and seems to act as a blinking, chirping second set of eyes on the road (or a backseat driver, if you will). Unfortunately -- due to garbled machine translation -- it's nearly impossible for us to discern all the features offered by this device, but we think that its embedded camera and image recognition software allow it to remind you when the light has changed, alert you if you're swerving ("Officer, I swear I'm not drunk -- see, my robot penguin hasn't chirped at me once"), or give you a heads-up when you pass a "point of interest" (such as a ferris wheel, helpfully suggests the Impress write-up). No word yet on pricing or availability for this lively road trip partner, nor is it clear whether sporting one qualifies you for the carpool lane. Keep reading to check out the unnamed bot "in action" (i.e. sitting motionless in front of a driving simulator), and then hit the read link for a few more snaps and short video...
Search By Camera! delivers product data from cellphone pics
Much like the Photo2Search project being developed by Microsoft Research Asia, a new cameraphone-based image recognition service called Search By Camera! is promising to help consumers acquire info on products by simply snapping a picture while shopping. Developed by Bandai Networks and D2 Communications, using technology from Evolution Robotics (which is also at the core of a similar service launched earlier this year, called Mobot Mobile Visual Search, as well as having been recently licensed by Robosapien manufacturer WowWee Robotics), Search By Camera! will currently only work for the handful of folks who happen to have imported a DoCoMo FOMA N902iS phone into the US. Further degrading its utility is the fact that Bandai and D2C must get permission for each and every product they want to include in their database, which has so far only come from a scant ten companies. Still, even if this particular initiative doesn't take off, it's pretty clear that image recognition will play a leading role in the future of mobile search, with many handsets already possessing the proper tools that make such a service feasible.
