Google Maps has definitely improved over the years. Earlier this year, the app version got more service providers and full Uber service integration. The service added wheelchair-friendly information, parking difficulty icon for 25 metro areas in the US, real-time information, and even local reviews in your default language. The tech giant will not stop improving the web feature because it provides a lot of information to all the users.
Google Maps isn’t just a collection of maps from all over the world. It has become a useful tool for everyone. It’s not yet perfect but it is generally reliable. The company aims to deliver more accurate and timely information so it is adding more data. One of the many steps is to find a way on how to extract data automatically from billions of images collected by Street View cars. It sounds impossible but Google has already started with the work.
The Google Ground Truth team has began its ‘Attention-based Extraction of Structured Information from Street View Imagery’ project with the aim of extracting street names, numbers, and businesses. This will hopefully keep Maps up-to-date all the time. Google is now turning to deep neural networks for this automation. The algorithm should be able to “read” content of images much easier now.
The model is now available for public use on GitHub. Only the devs and coders can understand what Google is trying to say but simply put, the company is doing everyhing in its power to become more intelligent and more useful.
Data gathered by Street View cars will be analyzed carefully and together with Deep Learning, the system will be able to identify more details like numbers and street names. With all these enhancements, Google can then map a whole town or city more effectively than ever.
SOURCE: Google Research Blog
