Self-driving cars can see better than humans can. With cameras and sensors, such as LIDAR, SONAR and
RADAR, cars can “see” frequencies that humans cannot. Cars can also measure distances with incredible precision. The amount of data collected and processed exceeds 4TB of data per car per day.
Unfortunately, seeing is not enough for safe operation. Cars need to understand what it is they are seeing and make decisions about what to do in real-time and without depending upon Internet access.
In an emergency situation, a vehicle needs to know if one obstacle is a person and another is not in order to decide to miss a collision with the person.
Neurala’s Brains for Bots software is highly efficient software that recognizes and finds objects in a moving vehicle using an ordinary camera and in real-time (10-20 fps, depending on hardware).
While the car is driving, the software can quickly provide labels for objects that the car sees as they move. The Neurala brain can quickly find people, cars, trucks, bicyclists and other objects in the video stream. This information can be coordinated with information from other sensors to produce a fast, accurate result without the need to contact the Internet.
During product development and after data is collected, Neurala software can be used to provide ground truth for tuning sensors. If camera data is collected at the same time as when data is collected from other sensors, Neurala’s software analysis can provide detailed information about what is seen in the environment.