Many industry experts have prescribed driverless technology as the antidote to the persistence of driving-related deaths or injuries. Elon Musk himself has said it’s the industry’s moral duty to develop and release autonomous technology for the masses as quickly as possible.
But in the process of developing technology that would enable autonomous driving, some startups are also working on a more immediate solution.
Nauto, a company in which Andy Rubin’s Playground Studios led a $12 million round, has rolled out a device that will detect when a driver is distracted and then warn the driver in real time when in high-risk situations. Otherwise, the company will note every instance of distracted driving — to which studies attribute 68 percent of car accidents — and compile a progress report or dashboard the driver can access, presumably while not driving.
The idea is, these various methods of feedback will help in training professional drivers to become safer, which in turn could help bring down their commercial insurance rates and avoid accidents. The miles driven by those professional drivers and the resulting data set could then be used to inform and ultimately serve as the brains for an autonomous-driving system — which Nauto hopes to have a part in developing — in the future.
Nauto’s second-generation device is an aftermarket kit that has inward- and outward-facing cameras that can detect whether the driver is distracted (by capturing images of the driver’s eyes, chest and head) as well as what is happening on the road.
Using a deep-learning algorithm, the system is able to detect when a driver is not paying attention by capturing images of what a distracted driver looks like and assessing it against a non-distracted control image.
“If you’re looking down at your cellphone, by looking at your head, chest and eyes I can detect you’re not looking at the road and you’re looking at something else,” Nauto CEO Stefan Heck told Recode. “A lot of technologies that [exist today] just track the eyes and that hasn’t been that accurate.”
Eventually, Heck says the device will be able to detect cognitive distraction or when a driver is simply spaced out — which is a lot harder to discern.
“You can’t just look at a single picture, they will look normal,” Heck said. “In that case, you have to watch over time that they’re not moving their head, not scanning their eyes, are standing still. If you watch over time it becomes very clear that they’re just sort of sitting there.”
That information could then be used in the case of an accident to ensure an insurance claim is accurate. For instance, the accident footage could show that the driver with the device was paying attention and it was the other car’s fault.
Some insurance companies have been incentivizing professional drivers to use the device. This quality of information could help the insurance industry reimagine how policies work in the age of autonomous cars, something on which Nauto has already begun discussions.
So far, the company has retrofitted its device in dozens of commercial fleets and is also working with multiple automakers, including Toyota and BMW.
“We’re deploying into a bunch of different automakers’ fleets on both development cars and car-sharing fleets, and we’re already live in a couple of those,” Heck said.
Down the road, the company plans to integrate into vehicles at the production level. The deep-learning-backed camera system and feedback mechanism as well as the miles driven using the device could prove to be valuable for automakers attempting to develop safe driving systems and eventually driverless cars.
This article originally appeared on Recode.net.