Google Goggles Review

@Tim | Search

Officially, I found out about the Google Goggles release on the MSNBC report last week along with the rest of the world.  Unofficially, I’ve known Google has been investing heavily in their image search technology for about 18 months.  In August of 2006, Google acquired a tiny company called Neven Visions for an undisclosed sum.  With this small tactical acquisition, Google acquired a handful of interesting patents around image search and computer vision, as well as the scientific talent from MIT, Stanford and the like.  I did some digging and found some of the underlying patents, applications and abstracts:

Method and system for customizing facial feature tracking using precise landmark finding on a neutral face image

Image base inquiry system for search engines for mobile telephones with integrated camera

Image-based search engine for mobile phones with camera

Single image based multi-biometric system and method

Since the acquisition of Neven Visions, there have been a number of start-ups in the image search niche such as, Snaptell (acquired by A9/Amazon.comCo-Founder Don Tanguay now at Google Images), Linkme Mobile, Kooaba, and Pongr.  As these start ups had/have very limited resources in terms of indexing capabilities and infrastructure, becoming the default image search engine to compete with Google or Bing is next to impossible.

While each of these companies may have comparable or better technology as Google Goggles, I believe that none of them have a viable business model without the right distribution/infrastructure partner and foresee companies such as, Bing, Facebook, Baidu, Twitter etc to target these small startups to put their stake in the Augmented Reality space.

What are the limitations of Google Goggles?  Thanks to Google’s own image index, Google Goggles has a virtually unlimited potential image matching opportunities, but at this point I can’t seem to find any information on exactly what the real # of truly indexed or “Goggle Enabled” images are (my guess is around 3-8 million).  Google Goggles seems to have the same limitations as all other leading image search technologies such as, limited 3-D matching, clothing, blur, light, distortion, angled pictures etc.  Logos seem to work quite well from packaging, advertisements, and other flat surfaces, but it doesn’t work near as well for 3-D outdoor signage.  I’m actually surprised they put wine as a category; as vineyard labels are virtually the same for pinot noir, cab sav etc, which could lead to an increase in false positives.  Either way, a marriage between robust computer vision and the virtually unlimited Google image resources is a huge step in the right direction for Augmented Reality.

So what is the next step for Google Goggles?  My guess is the world’s most robust Augmented Reality browser and the first to bring together image search and LBS using Google Image and Google Maps as the overlays.  I’m sure the guys at Layar’s and Wikitude are playing the Google Goggles announcement as a positive, but I still find it very hard to imagine how any of these competing Augmented Reality browsers can ever deliver such robust image recognition.  While Layar has done particularly well (in terms of buzz not revenue) over the past 6 months due to the openness of their api, I expect Google Goggles would have a similar extension developer network for Goggles as they have for the many other products allowing developers to create their own plug-ins.