Semiconductor supplier Renesas Electronics Corporation has announced an agreement with StradVision, Inc., a vision processing technology solutions provider, to jointly develop a deep learning-based object recognition solution for smart cameras used in next-generation advanced driver assistance system (ADAS) applications and cameras for ADAS Level 2 and above autonomous vehicles.
StradVision’s deep learning–based object recognition software has been optimized for Renesas R-Car automotive system-on-chip (SoC) products R-Car V3H and R-Car V3M, which have an established track record in mass-produced vehicles. These R-Car devices incorporate a dedicated engine for deep learning processing called CNN-IP (Convolution Neural Network Intellectual Property), enabling them to run StradVision’s SVNet automotive deep learning network at high speed with minimal power consumption. The object recognition solution realizes deep learning–based object recognition while maintaining low power consumption, suiting it for mass-produced vehicles and encouraging ADAS adoption.
StradVision’s SVNet deep learning software is highly regarded for its recognition precision in low-light environments and its ability to deal with occlusion when objects are partially hidden by other objects. The basic software package for the R-Car V3H performs simultaneous vehicles, person and lane recognition, processing the image data at a rate of 25 frames per second, enabling swift evaluation and POC development. Using these capabilities as a basis, developers can customize the software with the addition of signs, markings and other objects as recognition targets.
In addition to the CNN-IP dedicated deep learning module, the Renesas R-Car V3H and R-Car V3M feature the IMP-X5 image recognition engine. Combining deep learning-based complex object recognition and highly verifiable image recognition processing with man-made rules allows designers to build a robust system. In addition, the on-chip image signal processor (ISP) is designed to convert sensor signals for image rendering and recognition processing, making it possible to configure a system using inexpensive cameras without built-in ISPs, reducing the overall bill-of-materials (BOM) cost.