A New System Lets Self-Driving Cars “Learn” Streets On The Fly

A New System Lets Self-Driving Cars “Learn” Streets On The Fly

University of Cambridge has made a fresh, new system known as SegNet which is able to examine roads as well as access different functions such as road markers, street indicators, pedestrians and even the sky.

SegNet takes a close look at any RBG image of the road before classifying various layers of it with the help of Bayesian analysis.


This is what the release states:

The first system, called SegNet, can take an image of a street scene it hasn’t seen before and classify it, sorting objects into 12 different categories — such as roads, street signs, pedestrians, buildings and cyclists – in real time. It can deal with light, shadow and night-time environments, and currently labels more than 90% of pixels correctly. Previous systems using expensive laser or radar based sensors have not been able to reach this level of accuracy while operating in real time.

The latter portion enables any motor vehicle to orient itself regardless of its position.

That goes to show that the system may view an image, thereby, assessing where it is situated along with its orientation as narrow as a few degrees.

It beats GPS by those standards and requires no use of a wireless connection to carry out these functions.

Even you can give SegNet a go as you leave it to follow any road in the area where you reside.

It is capable of analyzing any images of roads that it comes across and will continue to inform you of what it can view.

The advantage of using such a system mainly lies in the fact that it is capable of eliminating GPS; it rather stresses on the concept of machine learning within 3D space.

Though, it is by no means impeccable as we speak.

Professor Roberto Cipolla, the research leader, is upbeat about the prospects of SegNet: “In the short term, we’re more likely to see this sort of system on a domestic robot – such as a robotic vacuum cleaner, for instance. It will take time before drivers can fully trust an autonomous car, but the more effective and accurate we can make these technologies, the closer we are to the widespread adoption of driverless cars and other types of autonomous robotics.”

Source: TechCrunch