Industry news--Karmin Electronic (HK) Limited

Renesa Electronics released RZ/V series devices with built-in visual AI accelerator to achieve accurate image recognition and multi camera image suppo
2022-10-21

The new RZ/V2MA MPU has an OpenCV accelerator for image processing and provides a TVM based deep learning compiler

On September 29, 2022, Rexa Electronics (TSE: 6723), a global semiconductor solution supplier, announced today that it would introduce a new RZ/V2MA device capable of processing image data from multiple cameras, further expand its RZ/V series microprocessor (MPU) product line, and bring a new level of high-precision image recognition capability to visual AI applications. The new device integrates two 64 bit Arms ® Cortex ®- The A53 core can provide high computing performance at the highest operating frequency of 1GHz; It has a proprietary low-power DRP-AI (dynamic reconfigurable processor) accelerator, which can process image data with a performance of 1 TOPS/W (millions of operations per second/watt).

The RZ/V2MA device is equipped with high-speed interfaces such as Ethernet, USB, and PCI Express, allowing images to be input from multiple external cameras. In addition to the DRP-AI accelerator, RZ/V2MA also includes an OpenCV accelerator to simultaneously perform rule-based image processing. These functions enable machine vision products such as AI gateways, video servers, security doors, POS terminals and mechanical arms to obtain high-precision image recognition capabilities.

The new RZ/V2MA is equipped with a complete set of development tools to assist the design of visual AI system. In addition to the existing DRP-AI Translator, the new device also adds DRP-AI TVM (Note 2) based on the open source deep learning compiler Apache TVM technology (Note 1). The DRP AI Translator aims to convert AI models into DRP-AI executables, while the DRP-AI TVM compiler allows the DRP-AI accelerator to work with the CPU, enabling DRP-AI to convert and generate more AI models than ever before. Renesar provides support for ONNX and PyTorch AI models in the first stage of industrial development, and plans to support Tensorflow in the future.

Maoshu Kato, Vice President of Enterprise Infrastructure Business Department of Renesa Electronics, said: "For embedded system developers who want to implement machine learning, one of the challenges they face is how to keep up with the latest AI models that are constantly evolving. With the new DRP-AI TVM tool, Renesa brings designers the choice to expand AI frameworks and AI models. These frameworks and models can be converted into executable formats, enabling them to use the new AI models to build cutting-edge image recognition functions for embedded devices."

Chiharu Nakabayashi, president of amnimo Inc., a wholly-owned subsidiary of Yokogawa Electric, a service provider based on the Internet of Things and AI, said: "Rexa RZ/V series is an ideal choice for embedded devices. It only generates extremely low power consumption and heat when running AI, so it does not need fans or radiators. With these devices, we are confident to develop a powerful image AI gateway that can be installed in any occasion."

Key characteristics of RZ/V2MA
Ÿ Dual 64 bit Arm Cortex-A53 core, with maximum operating frequency of 1GHz
Ÿ AI accelerator DRP-AI (1 TOPS/W level), reaching 52 fps (frames per second) when executing TinyYoloV3 program
Ÿ OpenCV accelerator for rule-based image processing
Ÿ Ethernet, USB and PCI Express interfaces for external camera image input
Ÿ Video codecs (H.265 and H.264)
Ÿ DRP-AI TVM tool, used to transform AI models based on TVM technology; At present, ONNX and PyTorch formats are supported
Ÿ High speed memory interfaces include LPDDR4 (3200Mbps), USB 3.1 (up to 5Gbps), and PCI Express (2-channel)
Ÿ Available in 15mm square BGA package

Visual AI Gateway Solution Based on RZ/V2MA

As an AI based object detection and recognition platform, the "Visual AI Gateway Solution" developed by Renesa uses multiple cameras to collect data and obtain effective wireless data transmission. This high-speed processing solution combines RZ/V2MA MPU with power IC, VersaClock clock generator, and communication modules for Wi Fi, Bluetooth, and LTE. It not only brings flexible connection options and an optimized power system, but also has been tested to speed up the development of powerful AI gateway devices. As a part of Rexa's "successful product portfolio", this solution perfectly integrates compatible Rexa devices to achieve seamless collaboration, so as to help users shorten the design process and accelerate the product launch. Based on several products in its product portfolio, Renesa has launched more than 300 "successful product portfolio".

Supply information

RZ/V2MA and development tools are now available.

The amnimo website also provides demonstration videos of the AI edge gateway; Please visit: https://youtu.be/OUpY2w0VdNk 。

Prev：RAC15-K/480 and RAC25-K/480 of RECOM conform to OVC III and PD3 grade Next：The second generation product of Nidec drive motor system "E-Axle" starts mass production