Home Blogs Google uses Deep Learning with Street View to update its maps

Google uses Deep Learning with Street View to update its maps

Google has found a new way to update Google Maps, by combining deep learning with Street View. The process of updating the maps is both intelligent and interesting.

Using “Attention-based Extraction of Structured Information from Street View Imagery”, Google is implying deep learning on millions of daily collected images from Google Street View. The complex algorithm is based on a deep neural network that mainly focuses on extracting text from scanned images, leaving out on every other information that can be considered sensitive from a privacy standpoint.

1
Google Street View offers panoramic views, collects real-world- imagery to complement Google Maps and Google Earth.

How Street View works?

Google Street View, which offers 360-degree camera panoramic views, collects real-world- imagery to complement Google Maps and Google Earth. While the camera collects images, the algorithm based network works on extracting additional information from the photos — including street numbers and names — to improve the data available in Google Maps.

But when it came down to accuracy, it showed a sequence error rate of 15.8%, which could have been a substantive failure for Google, because most of us rely on Google Maps for their accuracy. However, after analyzing the failure cases, it was found that 48% of those error cases were because of the ground truth errors. This was enough to highlight the fact that the deep neural network model is on par with the label quality.

How Deep Learning method can help?

The method could have been used to improve location data in one-third of addresses globally. But since addresses don’t just consist of numbers, which is why Google has been working to expand the system to include street names, too. The system is also capable of replacing abbreviations with full names (e.g. “Avenida” rather than “Av.”) and ignoring any irrelevant text within the photo.

Using the automated deep learning system, Google can enhance its ability to map new streets and add buildings that may not yet have made it onto a city’s official maps. And not just that, with some new mesh, the same system can help Google read business names, which can enable it to automatically update new business listings on Google Maps.

So by combining the location data from Street View car’s GPS with address information and business names extracted from imagery, Google could effectively map an entire city without any pre-existing knowledge of the layout or nomenclature.