#SMTLive today!

Google Develops Machine Learning System to Recognize Business Names from Street View Images

Registering your business details via Google My Business is a key step in the SEO process. Logging your business details enables Google to better index the relevant information, and present it in related search results, which can have significant benefits.

Google Develops Machine Learning System to Recognize Business Names from Street View Images | Social Media TodayAnd now Google could be close to making this process even easier, requiring less input from business owners. According to a new report on the Google Research blog, they’re working on a new system that’s able to recognize business names and addresses via Google Street View images, which are constantly updated to ensure they reflect the latest info.

As per the report, Google has developed a new process which is able to identify the text on street signs from ‘very challenging’ Street View images.

Google Develops Machine Learning System to Recognize Business Names from Street View Images | Social Media TodayAccording to Google, their algorithm now achieves “84.2% accuracy on the challenging French Street Name Signs (FSNS) dataset, significantly outperforming the previous state-of-the-art systems” – results which are good enough for them to release the model for wider use. But while the focus is on street signs, and how they can help Google ensure the accuracy of their Maps listings, consistent with real-world experiences, the other key element is businesses.

As noted by Google:

“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.”

The system is still in development, but Google explains that in their work on the process (which has been in development for the last few years), they’ve been able to overcome various challenges in object recognition to improve the accuracy of their listings, including the translation of relevant data from storefront signs.

“Once a storefront is detected, one still needs to accurately extract its name for it to be useful - the model must figure out which text is the business name, and which text is not relevant. We call this extracting “structured text” information out of imagery. It is not just text, it is text with semantic meaning attached to it.” 

Google Develops Machine Learning System to Recognize Business Names from Street View Images | Social Media TodayIn this image, Google says their system was able to accurately identify the business name ‘Zelina Pneus’ despite the accompanying tire brand signage that could confuse the model.

In future, Google says that this system will be able to accurately 'read' business names, enabling them to cross-check that info with their existing system data to provide more accurate and up-to-date business listings.

It’s another example of how machine learning and big data are changing the discovery process, and simplifying on and offline connection. Google’s Street View cars collect millions of images every day, building upon Google’s data resource of more than 80 billion high resolution images. Of course, analyzing data at that scale requires a lot of computing power, but Google's moving closer to making this a reality. There may come a time where Google has most of the data it needs on your store front before you even enter it, helping them to improve the accuracy of their knowledge graph, geographic targeting, ad delivery, etc.

Really, we’re only just now beginning to touch upon the possibilities of such technology – it’ll be interesting to see how such data and insight can be used in future to improve the user experience. 

Join The Conversation

Webinars On Demand

  • May 09, 2017
    With all of the technologies available to marketers today, have we lost that personal touch? Join VP of Content Marketing for ON24, Mark Bornste...
  • April 05, 2017
    In the ever-changing world of digital marketing, operational efficiency, quick turn-around times, testing and adapting to change are crucial to...

Whitepapers