Google details how it makes maps better with new AI features

  • by
  • 3 min read

At Google I/O, the search big outlined two new features of Google Maps that may enhance the pace of mapping by greater than ten instances in comparison with 5 years in the past. This new characteristic is designed to deliver maps containing helpful info to customers, and Google is detailing how AI makes these two new features doable. One technique is to make use of AI to show maps to acknowledge and predict when individuals will step on the brakes.

Features AI is getting used to tell drivers when to use the brakes as a part of its route replace, which may help customers keep away from conditions the place they should apply the brakes, comparable to chaotic lane adjustments or freeway exits. Google is utilizing synthetic intelligence and navigation info to establish braking occasions, that are moments that trigger the driving force to decelerate shortly. Hard braking is a recognized indicator of the probability of a automobile crash.

Google mentioned it believes that this replace will assist get rid of greater than 100 million emergency braking incidents on routes pushed on Google Maps every year. Google skilled AI on two knowledge units utilizing machine studying fashions. The first knowledge set accommodates telephone info utilizing Google Maps, as a result of telephone sensors can decide the diploma of deceleration alongside the route. Data is vulnerable to false alarms, as a result of the telephone can transfer within the automobile and has nothing to do with the automobile itself.

To remedy this downside, when utilizing Android Auto to venture on the show, Google will use the knowledge from the route pushed by Google Maps. The search big admits that it is a comparatively small subset of information, however its accuracy is excessive as a result of Maps is in a secure place. Google can also be working to seek out different causes for the arduous slowdown, comparable to structural or visibility points.

Google additionally confirmed that by the tip of 2021, detailed avenue maps can be expanded to greater than 50 cities. Again, use synthetic intelligence to realize this objective. This requires Google to remake its mapping course of. Its well-trained machine studying mannequin can acknowledge and classify numerous features, together with hundreds of thousands of avenue scenes, satellite tv for pc and aerial photos from the start of the street to shifting to buildings and different features. Its mannequin has been up to date to acknowledge all objects within the scene without delay, in order that it can detect and classify a variety of characteristic units whereas sustaining accuracy, thereby dashing up map manufacturing.