Latest RZ/V2MA Microprocessor From Renesas Features Acceleration Engines For OpenCV And Deep Learning

Renesas Electronics Corporation offers complete semiconductor solutions enabling billions of connected, intelligent devices to improve people’s work and life. These solutions are based on trusted embedded design innovation. Renesas, a market leader in microcontrollers, analog, power, and SoC technologies, offers complete solutions for various automotive, industrial, home electronics, office automation, and information communication technology applications that contribute to the creation of an unbounded future.

Renesas has unveiled a new chip called the RZ/V2MA microprocessor, designed with the Apache TVM compiler stack and intended to speed up OpenCV and other machine learning workloads for low-power computer vision at the edge.

Two 64-bit Arm Cortex-A53 processor cores capable of operating up to 1GHz and a DRP-AI artificial intelligence coprocessor with a computing capacity of one trillion operations per second (TOPS) per watt make up the new Renesas RZ/V2MA. This translates to a 52-frame-per-second real-world performance for the TinyYoloV3 network.

The device has a DRP-AI ( Dynamically Reconfigurable Processor )coprocessor and an accelerator designed to speed up OpenCV workloads. Boosts performance for rule-based image processing, which can operate concurrently with networks powered by the DRP-AI.

The DRP-AI TVM from Renesas allows programs to be developed to operate on the DRP-AI accelerator and one or both CPU cores. It is built on the open-source Apache TVM deep learning compiler stack. The firm also provides the DRP-AI Translator software, which offers the conversion for ONNX and PyTorch models to run on the DRP-AI core, with TensorFlow support.

The DRP-AI TVM tool, according to Renesas partner amnimo, will provide “the most recent image recognition capabilities” for embedded devices. (Amni: mnimo)

According to the study, one of the difficulties for embedded systems engineers who want to use machine learning is keeping up with the most recent AI models, which are constantly evolving. Designers may now add the most modern image recognition capabilities to embedded devices using new AI models by expanding AI frameworks and models that can be translated to executable forms using the new DRP-AI TVM tool.

Additionally, the RZ/V2MA supports two lanes of PCI Express, up to 3.2Gbps of LPDDR4 memory, USB 3.1 connectivity, and the H.264 and H.265 video codecs. Renesas put the chip in the Vision AI Gateway Solution reference design, which incorporates Ethernet, Wi-Fi, Bluetooth Low Energy (BLE), and cellular LTE Cat-M1 connectivity, to demonstrate its capabilities.


🐝 Join the Fastest Growing AI Research Newsletter Read by Researchers from Google + NVIDIA + Meta + Stanford + MIT + Microsoft and many others...