QCRI and MIT joined hands to improve navigation through Global Positioning System or GPS. And, they dabbled with images from satellite to invent new model for the same. Basically, they tag features on the road, making maps more detailed, improving navigation.
Understanding the Model
Details help navigate unfamiliar areas better. Plus, they provide better warning to drivers regarding lanes, and whether or not they are merging or diverging. Furthermore, parking woes stand addressed with this new system as it informs people of spots available or not available. Thus, it allows for proper planning, and negotiation of traffic on city streets. In fact, experts believe that it is a boon in relief initiatives in disaster-struck regions.
However, it is worth pointing out that not only is the process of map detailing expensive but also time consuming. Thus, while Google works with camera-strapped vehicles, it ignores certain regions due to high costs associated, especially with map updation.
And, experts state that a worthy way forward is by unleashing AI (artificial intelligence) to enable better machine learning. Mainly, once machines learn understanding satellite images, automatic tagging may occur. Here, it is worth mentioning RoadTagger, technology that predicts lanes and road type via decoding of neural networks of architecture. Basically, this seems to be a good way to overcome the challenge of obstructions in view via satellites.
Hype or Real?
In a test on covered roads, RoadTagger came through. Mostly, data from some 20 cities paved way for accuracy evaluation. And, the result was a 77% accuracy in terms of lane numbers and a 93% accuracy in terms of types. Going forward, the platform is now witnessing research towards decoding bike lanes and parking spots in the city.
Some of the co-authors of the paper include graduate students, undergraduate students, and QCRI and CSAIL professors.