According to Cohen, the environment perception systems currently available on the market are not optimal in terms of both accuracy and price. The reliability of these systems also leaves a lot to be desired, says Cohen. The software developers from Or Yehuda want to eliminate these weaknesses by directly processing the data from the sensors without relying on the pre-processing integrated in the sensors. This means that the Vayavision system has to process much larger amounts of data, but it can also identify the relevant objects in the vehicle's environment with greater accuracy. To handle the he huge amounts of data, the software males use of a Deep Neural Network (DNN) running on Nvidia hardware. In addition to the DNN, the software also uses classical image processing algorithms, says Cohen. What’s more, the company also took functional safety into account during the development on several levels. "We can even use it to detect unclassified objects," said Cohen.
Vayavision's product is aimed at automotive OEMs and Tier Ones. Cohen said that in Europe and Asia, "between 5 and 10" customers had already been acquired, but without naming any names. "The negotiations are all under NDA," said Cohen.