In recent tests, Google’s algorithm was able to read French street signs with an 84 percent success rate. What that means is that those Street View cars you see driving around may not only take photos for Street View, but they may also fill in Google Maps profiles automatically.
Of course, there are a number of factors that go into whether or not the algorithm can read the signs. Things like lighting, angles, cluttered backgrounds, and more, all affect how well the cameras can see signs. Still, an 84 percent success rate is far better than any previous algorithms, according to Google.
“Our algorithm achieves 84.2 percent accuracy on the challenging French Street Name Signs (FSNS) dataset, significantly outperforming the previous state-of-the-art systems,” said Google in a blog post. “Importantly, our system is easily extensible to extract other types of information out of Street View images as well, and now helps us automatically extract business names from store fronts.”
Google does already use neural networks in Street View. For example, as you browse through Street View you may notice that faces and number plates are blurred — which is the result of machine learning. The algorithms also use artificial intelligence to extract things like street numbers, which helps improve location data related to the images. Of course, numbers aren’t the be all and end all of location — which is why new algorithms will be able to read street names, too, using the same system that can read business names.
It will be interesting to see how Maps and Street View improve thanks to new algorithms like this, and Google will continue to make improvements to its machine learning algorithms.
Editors' Recommendations
- The 9 best new features in iOS 14: Beautiful widgets, cycling directions, more
- Believe it or not, the video phone turns 50 today
- How to use Google Maps
- A literal minority report: Examining the algorithmic bias of predictive policing
- How A.I. created that amazing sports highlight reel you can’t stop watching
