As mobile phones continue to develop, the improvements to geolocation features, video capabilities, and processor speed combined with APIs from various web services are helping to make augmented reality the next big thing in mobile applications. On open platforms like Android, we've already seen amazing developments, including things like Layar, an augmented reality app that describes the world around you, and TwittARound, an app which shows you nearby tweets. But one of the items on our Augmented Reality wishlist - AU facial recognition - isn't something we've seen come about just yet. It almost seems too futuristic to be real. And yet...here it is. Swedish software and design company The Astonishing Tribe is developing an AU concept called Augmented ID that "sees" people and tells you who they are.
AugmentedID uses facial recognition and tracking technology from a company called Polar Rose, a photo tagging startup. A few months ago, Polar Rose made the news when they released a tool that lets you tag your friends on flickr and then alert them that they were tagged via Facebook Connect integration. Although hyped as "facial recognition for flickr," the technology isn't a part of the photo-sharing site at all (although if Flickr was smart, they'd soon find a way to make that happen). Instead, Polar Rose, at its very core, is simply a facial recognition algorithm which pulls in photos from flickr to its site for tagging. It can also be integrated into any web site you own via a javascript widget. Apparently, it can now be integrated into mobile devices, too.

Polar Rose performs facial recognition by matching 3D models of faces, one of the three methodologies for performing this task. This technique captures the shape of the face and other distinctive features like the contour of the eyes, the nose, and the chin. The advantage of using this technique is that it's not affected by low lighting scenarios - a useful feature when used in real-world mobile interfaces like that of Augmented ID.
Unfortunately, in order for Augmented ID to know a person's face to begin with, it appears you first have to set up an Augmented ID account. Once that's done, you can maintain different profiles for your public persona versus your private one, switching between them as desired. Each profile has associated with it your contact information, web links, and information about your social networking profiles. Another Augmented ID user could then aim their phone at you and Augmented ID would recognize who you were and show them the appropriate profile - which is actually the video stream of you in real life with the various links sort of hovering around your head.
This isn't exactly the ideal mobile facial recognition solution just yet since it requires both participants to use Augmented ID in order for it to work. But it is one step closer than anything else we've seen before. In a perfect world, however, anyone could use the app to identify anyone else - regardless of whether or not that person had created an Augmented ID profile.
Of course, implementing that type of feature would be difficult, but not impossible. Given the numerous public photos on the web, an app could, in theory, use its facial recognition technology to compare the 3D model of the face in the live video feed to those out there on the web as opposed to ones in its own database. A good place to start this matching process would be Facebook. Despite the millions upon millions of photos on that social networking site, facial recognition already exists there courtesy of the new app Photo Finder. Perhaps the two companies could even work together so the mobile app could query against the photos Photo Finder has already turned into 3D models for matching purposes. (Is that technically possible? If that's your area of expertise, let us know).
That's getting a little ahead of what Augmented ID actually does, but possibilities like this are exciting.
At the moment, Augmented ID is more of a mobile "concept" than it is a working application ready for download. Still, it's only a matter of time before technology like this is adopted and put into use in the real world. In fact, that's why TAT developed it in the first place. The Swedish company designs and builds products and services that "enhance the user experience of portable devices." They work with OEMs including SonyEricsson, Motorola, S60, Samsung, Vodafone, and Orange as well as partners like Texas Instruments, Freescale, Teleca, Macnica Networks, Montavista, Nvidia, and Symbian. In other words, you may one day see this technology come built into your mobile device. But we'd settle for the Augmented ID mobile app -wouldn't you?
Image credit: engadget
Comments
Subscribe to comments for this post OR Subscribe to comments for all ReadWriteWeb posts
This is so cool!
I can't wait to install one for my phone.
The book of revelation to come true verbatim!
Posted by: Matthew DeVries
|
July 10, 2009 8:17 AM
Cool! Remembers me on ubiquitous social networks: http://tiny.cc/Y6okg
It was not possible to comment on TAT's site so I'll put it here. We brainstormed that thing recently. Especially along the lines of privacy and security.
Firstly a nice add–on to this feature would be "personal privacy firewall" where you select who, what and when sees something attached to your persona image. Imagine someone uses a photo instead of real face — this way I can track and find out other profiles which an original user was not expecting to be known to everyone. Of course it all brings us back to not to publish anything on the web that may harm yourself.
Secondly a "call log" feature would be nice.
Thirdly geo–fencing could certainly ease the process of switching profiles. How often, do you think, a person will forget to switch from "hand-out" to "business" profile after a wild Sunday party?
Fourthly "call log" plus social networking or micro blogging would be nice feature. Messages on twit* like "I just met …" or "Sitting in Starb* with *" (turning geo–location on).
Exactly as you say Valentin, we choose to let people choose what should be exposed about them. Possibly you could allow "public" info, such as name to be available, but I am not sure - don't want to create the ultimate stalker app, do we :)
/Hampus, TAT-founder
@Hampus great to see TAT people here! Love your company's blog on innovations — almost every post pinned to the wall waiting to be brainstormed :)
Talking about privacy. The way we saw it the problem was that client relies on only two parameters to present info: a face match and current selected profile. If it is so and the goal is to get as much profile information as possible then potential "privacy attacks" were either to trick the face matching or the "current" moment. This is why the four points I had were about covering these vectors.
what a load of streaming cock spunk
This is kinda what I hoped Bluetooth would be used for. I mean, holding up a camera to ID people is kinda rude and obvious. I want a standardised BT protocol for social 'real world' networking.
Build a bit of Palm WebOS Synergy so I can track people events for people I know and I become a social chameleon - remembering birthdays, kids names etc.
Course, Bluetooth, as we know, is for pushing adverts...
1.Nikia mobile with long-lasting battery and durable designs will help you conduct business in confidence and style.
I will definitely spread the word, my friends who are more into this thing would love this, thanks for the post.
Bluetooth an option I guess but what about users carrying their web identity using QR codes /bar code system? Rgds, yan
Matt - Bluetooth won't work if you or the person you are looking at is moving, ditto for wifi. Range is of course an issue for BT, as well as WiFi, as is latency (how many seconds does it take for your car's handsfree mode to connect?) The ideal technology for this app is DASH7, which is extremely low power, tracks moving objects, is already an important part of many "internet of things" visions, has multi-km read ranges, and has leading silicon providers like TI, ST Micro, ADI, Melexis, and Semtech developing single chip solutions for DASH7. I blogged about this very issue recently here: http://ow.ly/1aY1T
Yan - barcodes only work if you are VERY close to the object being measured/sensed. Great technology, but not ideal for this app IMHO.
Pat
Posted by: Pat Burns
|
February 24, 2010 2:34 PM