Renesas is expanding its RZ/V series of AI-enabled microprocessors (MPUs) with a device for processing image data from multiple cameras with high-precision image recognition for AI applications. The new device has two 64-bit Arm Cortex A53 cores and delivers high computing power with a maximum operating frequency of 1 GHz. RZ/V2MA also has a proprietary low-power DRP AI accelerator (Dynamically Reconfigurable Processor) for AI of class 1 TOPS/W (tera operations per second, per watt). It is available in a 15 x 15 mm BGA package.
RZ/V2MA includes interfaces such as Ethernet, USB and PCI Express for multiple cameras. In addition to the DRP AI accelerator, it includes an OpenCV accelerator for rule-based image processing. These functions provide image recognition for machine vision products such as gateways, video servers, security gates, POS terminals and robotic arms.
RZ/V2MA offers a suite of development tools for vision AI systems. In addition to DRP-AI Translator, it includes DRP-AI TVM based on the open source deep learning compiler in Apache TVM technology. The DRP-AI Translator converts AI models into executable DRP-AI files. The DRP-AI TVM compiler cooperates with the CPU via the DRP-AI accelerator. This allows DRP-AI to convert and generate more AI models than ever before. As a first step, Renesas supports ONNX and PyTorch AI models and plans to support Tensorflow.
"With the DRP-AI TVM tool, we offer the ability to extend AI frameworks and AI models that can be converted to executable formats," explains Shigeki Kato, VP of Renesas' Enterprise Infrastructure Business Division. "This allows the latest image recognition functions to be implemented in embedded applications."
Renesas has also developed the Vision AI Gateway Solution, an AI-based platform for object detection and recognition. It uses multiple cameras to efficiently transmit data wirelessly and combines the RZ/V2MA MPU with complementary products such as power ICs, VersaClock clock generator and communication modules for Wi-Fi, Bluetooth and LTE. The solution offers flexible connectivity options and has an optimized power supply. It has been tested to accelerate the development of robust AI gateway applications.
Hall: B4, Booth 141
