On Google I / O, the search giant describes two new features for Google Maps which helps map the road more than ten times faster than possible five years ago. This new feature aims to bring maps that are packed with useful information to consumers, and Google provides details about how AI makes two new features possible. One way is to use AI to teach maps to identify and estimate when people hit their brakes.
This feature utilizes AI to tell when drivers hit the brakes as part of the routing update that helps users avoid situations that require them to hit their brakes hard, such as changes in confusing paths or freeways. Google uses artificial intelligence and navigation information to identify loud braking events, which is a moment that results in the driver to slow down quickly. The hard braking is a known indicator of the possibility of a car accident.
Google says it believes this update will help eliminate more than 100 million hard braking events on routes encouraged using Google Maps every year. Google trains its AI using a machine learning model on two data sets. The first data set has information from the phone using Google Maps because the telephone sensor can determine the slowdown along the route. Data is susceptible to false alarms because the phone can move in the vehicle regardless of the car itself.
To defeat it, Google uses information from the route driven by Google Maps when projected on the screen using Android Auto. The search giant recognizes that it is a relatively small subset of data, but is very accurate because the map operates in a stable place. Google also works to identify other causes of hard deceleration, such as construction problems or visibility.
Google also confirms that the detailed road map will expand to 50 more cities at the end of 2021. Again, AI is being used to allow it. It requires Google to make the mapping process again. It has a trained machine learning model to identify and classify features in millions of road display, satellite, and air drawings that start with the road and then move to buildings and other features. The models have been updated to identify all objects in scenes at once, enabling detection and classification of broad feature sets simultaneously while maintaining accuracy, which leads to faster visposes.