By now, autonomous cars are on everyone’s radar, and some of the biggest names in both tech (such as Alphabet) and car manufacturing (such as Tesla and Audi) are working around the clock to get autonomous vehicles on the road.
But the development of self-driving vehicles has given birth to a wealth of secondary technologies, and companies to produce them. There are many lesser-known companies working on creating the ancillary technology required for autonomous cars to function – and most importantly, to function safely.
One example is Nexar, an Israeli company that has created a dashboard camera and accompanying app that uses machine vision to issue real-time warnings to drivers.
Nexar recently released a dataset comprised of the world’s largest collection of street pictures snapped from moving vehicles.
The dataset, known as NEXET, contains over 50,000 photographs from 80 countries showing road conditions in various states of weather and times of day.
Nudge in the right direction
Nexar is releasing the dataset for the expressed purpose of driving autonomous car technology forward. The photos are available to challengers in an open competition to create parts of an Advanced Driver Assistance System (ADAS) that would accurately detect hazards in a variety of conditions.
“We are releasing this dataset to you, our challengers, to empower you to build a truly smart collision prevention system that can work extremely well anywhere and at any time,” Nexar explains on their website.
One of Nexar’s goals is to generate as much data as possible.
Machine learning depends on having a large amount of diverse data, and currently, the autonomous car industry is limited because most data comes from “a small number of vehicles running in controlled environments or in simulation, which fail to perform adequately in real-world dangerous corner cases.”
Because passenger safety is at stake, autonomous cars can’t risk relying on incomplete data.
That’s why Nexar is using its products, and research challenges like the competition to create an ADAS, to crowdsource information from on-road driving scenarios.
Says Nexar, “The robustness of learning end-to-end driving policy models depends on having access to the largest possible training dataset.
This is just one of many examples of how tech companies and real-world drivers are contributing to the dream of self-driving vehicles.