Deep-Learning: Google Maps to become more accurate through Artificial Intelligence

May 8, 2017 | 2196 Views

Google Maps, the 'indispensable' navigation application owned by Alphabet Inc. will soon become more accurate, thanks to Machine Learning.

Google's proprietary Maps with options for navigation, business listings, public transport information, traffic updates is set to become more accurate owing to 'deep-learning' technology, which is being integrated into the app, a post on Google Blog said.

Google Maps and Street View are set to receive the deep-learning techniques to automate, 'mapping new addresses'. Many of you might have seen Google's Street view cars that travel around cities and towns taking photos of streets, shops etc. from its 360 degree camera.

Google then uses this captured photographic data to 'extract' information like street names, numbers to update Maps. Up until now, Google used crowdsourcing to help identify the content of distorted images through a program called ReCAPTCHA.

"Yet, the photos that the Street View cars take are not always focused the right way. There may be discrepancies in the angle, the lighting, distortions, cluttered backgrounds, which make it difficult for a person to decipher the names and numbers," Gizmodo reported.

However, it becomes physically impossible for people to analyse over 80 billion high resolution photos collected by the Street View cars to update information. As a result, the Ground Truth team at Google proposed to 'infuse' Artificial Intelligence and Deep Learning to extract information from geo-located images, the report added.

Google already uses machine learning to censor faces and car license plates. Now it is using the technology to extract information from street signs.

"Google stated it was able to improve location data of one-third of the world's addresses. The company improved the software to take not only the street numbers into account, but also the names. The algorithm is capable of replacing abbreviations with full names too and ignoring any irrelevant texts in the photos," the report said. Read More

Source: Times Now