Best of 2014: How Google Cracked House Number Identification in Street View
Google can identify and transcribe all the views it has of street numbers in France in less than an hour, thanks to a neural network that’s just as good as human operators. In January, its engineers revealed how they developed it.
Google Street View has become an essential part of the online mapping experience. It allows users to drop down to street level to see the local area in photographic detail.
But it’s also a useful resource for Google as well. The company uses the images to read house numbers and match them to their geolocation. This physically locates the position of each building in its database.
That’s particularly useful in places where street numbers are otherwise unavailable or places such as Japan and South Korea where streets are rarely numbered in chronological order but in other ways such as the order in which they were constructed, a system that makes many buildings impossibly hard to find, even for locals.
But the task of spotting and identifying these numbers is hugely time-consuming. Google’s street view cameras have recorded hundreds of millions of panoramic images that together contain tens of millions of house numbers. The task of searching these images manually to spot and identify the numbers is not one anybody could approach with relish.
So, naturally, Google has solved the problem by automating it.