Microsoft OneAlbum: Fusing Facial Recognition and Social Networks – The Future of Windows Live Photos?

Microsoft Research recently revealed one of their projects at TechFest 2010 which caught our attention – OneAlbum. TechFest is Microsoft’s annual showcase of emerging technologies from Microsoft Research and is a channel for these new incubation technologies to make their way into actual products. Here’s a description of what OneAlbum actually is:

OneAlbum is a project that uses face-recognition* and event-matching technology to retrieve automatically photos that would interest you from social networks and your friends’ online photo albums—and add them to yours.

OneAlbum automatically finds relevant photos in my friends’ albums on social networks or in shared albums, brings them to my album, and shows them side-by-side with the photos I’ve taken. The technology behind OneAlbum is a novel, unsupervised face-recognition algorithm. It analyzes the photos in my album to find automatically the faces of people I most care about, based on frequency of their appearance; no tagging is required. Then, using the social-network graph and other information, OneAlbum crawls my friend’s albums looking for photos of people that interest me. The algorithm was tested on real large-scale albums including tens of thousands of photos and achieved accuracy rates as high as 90 percent.

From the sounds of it, this technology seems perfectly aligned to what Windows Live Photos is trying to achieve – pulling together all the information on the web on what is important to you easily and hassle-free, and this is evident by looking at how Web Activities is already doing this. By combining face recognition technology and the web (note: face recognition is not face detection, it recognises the person in the photo and based on an algorithm it would automatically identify who it is), OneAlbum can pull together all the photos available on the web such as Facebook and Flickr, and group them automatically by person by automatically recognising who is in the photo. This pretty much eliminates the need for manual people tagging currently done in Facebook and Windows Live Photo Gallery! Here’s a screenshot (it even has a Windows Live header!):

7608.onealbum_thumb_10ee78d1 News

Not only facial recognition, but OneAlbum also involves event-matching technology by analysing the people, dates, colours, and textures in a set of photos and will recognise these photos as being taken in on the same event. Say goodbye to tagging your photos by events! You can check out a video of OneAlbum shown on Channel 9 here:

?LinkId=108181 News

Of course, one of the assumptions in this OneAlbum technology is that you must have an existing set of photos already uploaded (and possibly tagged properly first) for it to automatically match all those photos out there with your photos, and to figure out what is “relevant” and “most-interest” to you – a redefined concept called “My Albums”. Update: It is noted that OneAlbum works even if the albums or photos are not tagged. Surely this is an exciting piece of technology and we can’t wait to see it incorporated into Windows Live Photos or even Photo Gallery!

* According to Eyal Krupka, principal research program manager for Microsoft Israel R&D Center, the facial recognition technology used in OneAlbum is based on a collaboration between Microsoft Research Asia and Microsoft Live Labs. Perhaps this is already making its way into Windows Live Photo Gallery Wave 4?

Thanks to Picturepan2 @ LiveSino.net for the heads up!