Edge AI is becoming increasingly prevalent in embedded devices, particularly in those used in industrial and public facility cameras. This trend demands miniaturization and reduced power consumption in embedded products. Renesas Electronics Corporation has responded to this need by introducing the RZ/V2N microprocessor (MPU).
The RZ/V2N MPU incorporates an AI accelerator called Dynamically Reconfigurable Processor (DRP). The DRP-AI3, an evolution of this technology, offers improved AI inference performance. This advancement supports the growing demand for high-efficiency AI processing within a compact footprint.
Enhancing Efficiency and Compact Design
The RZ/V2N MPU is comparable to the high-end RZ/V2H MPU, boasting an AI accelerator, DRP-AI3, that delivers an impressive 10TOPS/W power efficiency. This efficiency is achieved through pruning technology, which boosts AI computational performance. Based on Resnet50, a representative CNN classification, the RZ/V2N provides an overwhelming performance benefit. It provides an improvement of 7 times more than the RZ/V2M without pruning (dense model) and 23 times more with pruning (90% pruning rate).
This high-power efficiency allows for extended battery life and the reduction of device size, as it eliminates the need for cooling fans.
Improved Image Recognition with Dual-Camera Input
The RZ/V2N supports dual-camera connections via its dual-channel MIPI-CSI interface, enhancing spatial recognition capabilities. This feature allows for higher-precision in applications such as flow analysis and fall detection.
For example, two cameras could be used to monitor the number of parked cars and identify their license plates on a single chip. In Driver Monitoring Systems (DMS), the RZ/V2N allows for monitoring of a vehicle’s interior and exterior using forward-facing and in-cabin cameras. The system can also gauge the distance to an object, improving image recognition accuracy in AI cameras and mobile robots.
Design Asset Reuse and Development Environment
The RZ/V2N shares a similar architecture with the RZ/V2H, utilizing a quad-core Arm® Cortex®-A55, a single-core Arm Cortex-M33, and an Image Signal Processor (ISP). This similarity simplifies software development by enabling the reuse of assets from the RZ/V2H. Extensive compatibility of peripheral functions allows users to select from a wide range of RZ/V series MPUs based on their AI performance, system configuration, and real-time control needs.
To facilitate quick evaluation and development, Renesas has released the RZ/V2N AI Software Development Kit. This kit comprises Renesas evaluation board kits, a software development environment, and over 50 AI applications tailored for various use cases. Furthermore, users with existing AI expertise can develop custom AI models and integrate them into the RZ/V2N MPUs.
Additionally, Renesas ecosystem partners will provide System on Module (SoM) boards, Single Board Computers (SBCs), and camera modules incorporating the RZ/V2N. This approach allows users to focus on developing AI applications, enabling rapid product development. Both the device and its evaluation board are available for purchase from the Renesas Online Store. Visit www.renesas.com/rzv2n for more details.
*This product partially utilizes the results of projects commissioned by the New Energy and Industrial Technology Development Organization (NEDO).
