Edge AI has been increasingly used in embedded devices such as industrial and public facility cameras, requiring miniaturization and low power consumption for embedded products. Renesas RZ/V series microprocessors (MPUs) are embedded with an AI accelerator – Dynamically Reconfigurable Processor (DRP) – known for its high-power efficiency that can meet the AI market needs. This DRP technology continues evolving into DRP-AI3, offering higher AI inference performance.
Extend Battery Life and Build Compact Products with High Power Performance MPUs
The RZ/V2N MPU, similar to the high-end RZ/V2H MPU (up to 80TOPS), is equipped with a high-power efficiency AI accelerator – DRP-AI3 – achieving AI performance of 10TOPS/W with pruning technology that enhances AI computational efficiency.
Looking at the following, Figure 1 compares the AI inference performance of the conventional RZ/V series with the RZ/V2N, RZ/V2N achieves an overwhelming performance of 7 times more than the RZ/V2M without pruning (dense model) and 23 times more with pruning (pruning rate 90%) in the case of Resnet50 which is a representative classification CNN. This extremely high-power performance enables extended battery life and miniaturization of devices, as no cooling fan is needed.
Dual-Camera Input Enhances AI-based Image Recognition Accuracy
RZ/V2N supports the connection of two cameras through its dual-channel MIPI-CSI interface (Figure 2). This improves spatial recognition performance compared to traditional single-camera AI image recognition, enabling high-precision applications such as flow analysis and fall detection. For example, dual-angle imaging with two cameras can be used to monitor the number of parked cars and recognize license plates on a single chip. In Driver Monitoring Systems (DMS), RZ/V2N enables monitoring and recording of both the interior and exterior of a vehicle using forward-facing and in-cabin cameras. Also, it is possible to estimate the distance to the object being imaged by configuring a stereo camera with two cameras, enhancing the accuracy of AI-based image recognition in AI cameras and mobile robots.
High Compatibility with the Same Architecture as RZ/V2H to Reuse Design Assets
The RZ/V2N is similar to the RZ/V2H, featuring a quad-core Arm® Cortex®-A55, a single-core Arm Cortex-M33, and an Image Signal Processor (ISP), allowing the reuse of software development assets from the RZ/V2H. Additionally, due to the high compatibility of peripheral functions, users can choose the optimal product from a wide range of RZ/V series MPUs depending on their AI performance, system configuration, and the necessity of real-time control.
To allow users without deep AI expertise to quickly evaluate and develop AI applications, Renesas released the RZ/V2N AI Software Development Kit that includes Renesas evaluation board kits and the software development environment along with over 50 various AI applications for different use cases. Use this development environment to reduce time to market. Additionally, users with existing AI knowledge can choose the AI compiler to develop their own AI models and integrate them into the RZ/V2N MPUs (Figure 3).
Furthermore, Renesas ecosystem partners will sequentially provide System on Module (SoM) boards, Single Board Computers (SBC), camera modules, and more that incorporate the RZ/V2N. This allows users to be free from hardware design and focus on AI application development, enabling rapid product development.
This unique embedded AI processor, RZ/V2N, is ideal for your next edge AI development. Both the device and its evaluation board (RZ/V2N-EVK) are available for sale 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).