Today, most of our searches on the internet land on an online map for directions, be it a restaurant, a store, a bus stand, or a clinic. It has become our virtual compass to finding our way through densely populated cities or even remote pathways. Google Maps is one of the most accurate and detailed maps available today. Ever wondered what makes it so accurate. Wondered how it knows the location of a new restaurant or the junction where there is a roadblock. Well, the technology behind this mastermind is Machine Learning (ML).
Machine Learning Algorithms
Machine Learning (ML) is a branch of AI that provides applications the ability to learn through experience. It uses statistical models and algorithms to perform tasks that are not explicitly programmed. To perform a task, Machine Learning Algorithms build a mathematical model based on sample data.
For instance, we can identify the faces of people around us such as our friends, family, and colleagues. However, if a programmer has to write a program to do what our mind does, it would be impossible to do so. This is where Machine Learning is more practical. You don’t need to tell the Machine Learning Algorithms what to do and how to do it. Instead, you provide the algorithm with lots of examples (in this case, image) and the algorithm learns from these examples.
The data gathered is also known as ‘training data’, which is used by the ML algorithms to build a statistical model that can solve problems through experience.
Artificial Intelligence and Machine Learning in Google Maps
Google Maps provides useful directions and real-time traffic information to millions of users. This information is updated constantly to mirror the changes in an ever-changing world. It is impossible to manually analyze more than 80 billion images to find new, or updated, information for Google Maps. One of the goals of Machine Learning is to enable the automatic extraction of information from geo-located imagery to improve Google Maps.
When it comes to accuracy, precision, and update frequency of maps, the pressure has never been higher. Robust training data is the backbone to develop machine learning algorithms. Machine learning and large-scale image collection are the keys to providing location data in a scalable way.
Images and Data
Satellite imagery has been an important part of identifying where places are in the world. It shows where roadways, open fields, buildings, and businesses are located in a region. Billions of images are collected from different countries and authoritative data is added to these images. The addition of data to these images brings the maps to life. Data is collected from various sources, like data from a local municipality, a housing developer, a business owner, or the local government.
How Machine Learning works in Google Maps?
Imagery and authoritative data are static and can’t keep up with the ever-changing world around us. Machine Learning algorithms can analyze existing images and data and identify changes in the new data. Thus, the maps are updated with only the recent changes. This increases the speed of mapping and allows for automation of mapping processes, while maintaining accuracy.
It makes use of a deep neural network that automates the image information reading process. This algorithm is publicly available on GitHub through TensorFlow, which is Google’s own open-source machine learning software library.
Google is already implementing machine learning to identify car license plates. And now, it is using the same technology to fetch information from street signs. Using this technology, Google aims to improve the location data of about one-third of the world’s addresses. The latest Machine Learning algorithms helped achieve 84.2 % accuracy when tested on several challenging street signs in France. These statistics significantly outperformed the previous state-of-the-art systems.
This move improved the software to read the street numbers and street names. The new algorithm can get rid of any irrelevant text in its photos and replace abbreviations with their full names.
Mapping Building Outlines in Google Maps
Previously, an algorithm that tried to guess whether part of an image was a building or not, resulted in amorphous blobs, that didn’t look like real buildings when you drew them on a map. Buildings are landmarks and a key part of how someone knows where they are when looking at a map. To fix this, the Google data operations team worked to trace common building outlines manually, and then used this information to teach the machine learning algorithms which images correspond with building edges and shapes. This technique proved effective, enabling Google to map as many buildings in one year as they mapped in the previous ten years.
Now when a new building or business comes up in an area, the Machine Learning algorithms identify the change and update the existing map, instead of remapping the entire area. This saves considerable time and effort.
Updating Real-time Transit Data in Google Maps
Google has come up with a new way to keep people informed, in real-time, about the status of their bus rides. Google Maps will has predictive capabilities – enabled through machine learning — to inform passengers well ahead of time if their buses are going to encounter some obstacles. It now provides real-time tracking data, which can forecast delays in hundreds of cities worldwide.
Google built a model that used standard traffic data, adjusted for the peculiarities of bus movements and routes. The team extracted training data from sequences of bus positions over time, as received from transit agencies’ real-time feeds, and aligned them to car traffic speeds on the bus’s path during the trip.
Finding Most Popular Dishes in Restaurants
Google Maps will help restaurant diners know what to order. An update to Google Maps will highlight the restaurant’s most popular dishes. This feature uses machine learning to uncover these dish suggestions based on the restaurants’ reviews and photos.
If diners have praised a dish in their review, Google Maps will match the dish to the photos uploaded by diners to create its selection of what’s popular there. Diners will be able to help by uploading photos of their meals on Google Maps.
The Overview tab in Google Maps will display the dish suggestions. When you see a dish you like, tap on it to see all the reviews where the dish is discussed by other diners.
These features that are built using Machine Learning have transformed Google Maps from a simple mapping utility to an application that offers a personalized experience to its end-users.
After reading about the Machine Learning Applications and Machine Learning algorithms implementation, are you ready to transition into your dream career? Consider Springboard’s AI/Machine Learning Career Track. Master foundations of data science and choose your focus area with a guaranteed job on graduation!