Netradyne, a computer vision startup for trucking fleets, has announced that its proprietary Driveri vision-based driver recognition safety program has captured and analyzed over 100 million minutes of driving video data. Unlike legacy companies and market incumbents who consider measuring the number of trucking miles being driven as a parameter for capturing data, Netradyne insists that monitoring driver behavior by the minute rather than the mile be ultimately more rewarding.
“One thing that we all know is that drivers put in long hours, with 8 or 9 hours a day, for 5 or 6 days a week throughout the year, averaging close to a 100,000 miles a year in some segments - whether it is ride sharing or delivery services, all the way up to the heavy duty freight segments. And our cameras are about observing the driver for all of those minutes of driving,” said Sandeep Pandya, the president of Netradyne.
Pandya explained that Netradyne is capturing data at a furious pace, with its AI-based computer vision models working in understanding vehicle behavior in traffic and driver behavior behind the wheel with concerns like seat belt compliance, and drowsy and distracted driving.
Netradyne’s primary camera system is powered by an NVIDIA TX1 teraflop processor, which Pandya mentioned to be the most powerful mobile GPU in the market at the time of the camera’s design. “The device is built with four HD cameras, with deep learning models and computer vision algorithms being the brains of the device and our secret sauce,” said Pandya. “This is what makes our camera system one of the most advanced in terms of the number of analytics and safety-related events that we see.”
The models are trained and optimized on millions of real-time roadway images, which helps it in identifying aberrant behavior and alerting drivers in real-time to avoid incidents on the road. “The real-time coach provides driver feedback on critical safety-related events, like running a stop sign or a light, or even if the driver is drowsy or distracted,” said Pandya. “Those are all things that in addition to sharing with the fleet manager, we can alert the driver ahead of time, as things that could help prevent a possible incident and reinforce it to the driver till they are conscious of the behavior that needs improvement.”
Netradyne is partnering with big players in the insurance industry, helping them analyze risks better, and in a way is working on veering the industry away from higher risks, by reducing the rate of incidents - and thereby the risk involved - through preventive analysis.
“When you look at the legacy video recorder providers, their products get triggered only with inertial events, while we’ve established that a driver can drive anywhere between 400-500 minutes a day. In terms of events and risks, the industry leaders are capturing less than half the percentage of the actual driver related behaviors at risk, and that is based on stats from their own websites,” said Pandya.
“Some of them capture 4-5 incidents a week and some 4-5 incidents a month. Each of those events represents 15-20 seconds of driving, and so in a 400 minute day, it is a minuscule percentage that you record as a coaching or safety tool. This is why we think the industry is ready for Netradyne - it is because people are really looking to see all of the risk involved.”
The value proposition of visual data analytics is enormous, as apart from fleets and insurance companies, the autonomous vehicle industry sees merit in understanding driver behavior, as at the core of its technology lies integration and co-existence with driver-driven vehicles on the road. “We are observing professional drivers on vehicles under all conditions and are profiling it in a way that no one else does in the industry,” said Pandya. Netradyne has recently raised $21 million in Series B financing and has a strong presence in San Diego and Bangalore, India with teams working on launching a driver safety platform for commercial vehicles in the near future.