Renesas Electronics has partnered with StradVision to develop the next generation of ADAS smart cameras

     Renesas has teamed up with StradVision, a provider of vision processing technology solutions for autonomous driving with deep learning expertise, to develop a deep learning-based vision processing solution for the next generation of smart cameras for advanced Driver assistance systems (ADAS) at ADAS Level 2 and above.

     To avoid common hazards that occur in urban areas, the next generation of ADAS will need to have high-precision visual processing capabilities that can detect so-called vulnerable road users (VRUs) such as pedestrians and cyclists. At the same time, for mid-to-low-end mass-market vehicles, these systems must achieve very low power consumption. The new solution from Renesas and StradVision achieves both goals and is designed to accelerate the widespread adoption of ADAS.

     Naoki Yoshida, Vice president, Customer Partnerships, Renesas Electronic Automotive Technology, said: "As a leader in vision processing technology, StradVision has extensive experience in developing ADAS solutions using Renesas R-Car SoCs. Through this collaboration, we are bringing production-grade solutions to deliver safe and precise mobile performance for the future. "This new deep learning-based collaborative solution, optimized for R-Car SOCs, will contribute to widespread adoption of the next generation of ADAS and support increasing vision sensor requirements in the coming years."

     Junhwan Kim, CEO of StradVision, said: "StradVision is excited to join forces with Renesas to help developers effectively advance the next big leap in ADAS. This joint effort will not only translate into a fast and effective evaluation program, but will also significantly improve ADAS performance. "The front-facing camera market is expected to grow significantly over the next few years, and StradVision and Renesas will no doubt maintain their leading position in this area by combining to create superior technology."

     StradVision's deep learning-based vision processing software provides high performance for vehicle, pedestrian and lane recognition. The high-precision vision processing software has been optimized for Renesas R-Car Automotive system-on-chip (SoC) products R-Car V3H and R-Car V3M, a family of products with a proven track record in production models. The R-Car family is powered by a CNN-IP (Convolutional Neural Network Intellectual Property) deep learning processing engine that enables it to run StradVision's SVNet automotive deep learning network at high speeds with the lowest power consumption. The vision processing solution developed in this collaboration enables deep learning-based vision processing capabilities while maintaining low power consumption, making it suitable for production vehicles and driving ADAS adoption.