What's that landmark? Google knows...

Tuesday 23 June

Google's admitted to researching computer vision, so how long before we see a visual search engine from the Big G?

Is there no stopping Google? Apparently the Big G has been researching “computer vision”. What does that mean exactly? That an online search function based entirely around images, with no text in sight, may not be that far off…

Google has already stepped up its research into the area and has already presented a new technology that enables computers to quickly and efficiently identify images of more than 50,000 landmarks from all over the world with 80 per cent accuracy. Okay, it’s not perfect, but it’s a start.

So, how does it work? Well, you start with an unnamed, untagged picture of a landmark, enter its web address into the recognition engine, and the computer will identify it and where it’s based, with not a hint of geo-tagging anywhere. It identifies the image from a database of 40 million GPS-tagged photos (from Picasa and Panoramio) and online tour guide webpages using image matching and clustering techniques.

It’s clever stuff. But we expect nothing less from the boffins at Google. Their aim is to unlock the information stored in the pixels of images on the web. The work has already been done for text, so why not for images? It’s still some way off, but if this system lifts off, then it could herald a whole new future for search engines…