Boost Energy Efficiency and AI Performance for Your Upcoming AI Cameras and Mobile Robots
Introduction to Edge AI and the RZ/V Series Microprocessors
Edge AI is becoming increasingly popular in embedded devices, particularly in industrial settings and public facilities where cameras are used. These devices need to be compact and consume very little power. The Renesas RZ/V series microprocessors (MPUs) are equipped with an AI accelerator known as the Dynamically Reconfigurable Processor (DRP). This technology is designed for high power efficiency, meeting the demands of the AI market. The latest advancement, known as DRP-AI3, aims to provide even greater AI inference performance.
Maximizing Battery Life and Compactness with RZ/V2N MPUs
The RZ/V2N MPU, like the more advanced RZ/V2H model (which offers performance up to 80 TOPS), comes with a high-efficiency AI accelerator called DRP-AI3. This technology can achieve AI performance levels of 10 TOPS per watt, utilizing pruning techniques to enhance computational efficiency.
An analysis of the performance metrics shows that the RZ/V2N significantly outperforms its predecessor, the RZ/V2M. In fact, it offers seven times the performance without pruning and up to 23 times better performance with a pruning rate of 90% when using ResNet50, a well-known convolutional neural network for classification tasks. This impressive capability means devices can be smaller and run longer on battery power since they do not require cooling fans.
Performance Comparison of RZ/V Series
- RZ/V2N Performance: 7x greater than RZ/V2M without pruning.
- RZ/V2N Performance: 23x greater than RZ/V2M with 90% pruning.
Enhancing Image Recognition with Dual-Camera Support
The RZ/V2N also supports the use of two cameras, thanks to its dual-channel MIPI-CSI interface. This feature significantly boosts the accuracy of spatial recognition when compared to devices using only one camera. Such capabilities are crucial for applications like flow analysis and fall detection.
For instance, by employing two cameras, one can monitor parking lots and even read license plates from different angles. In Driver Monitoring Systems (DMS), the RZ/V2N can effectively monitor both the interior and the exterior of vehicles by utilizing a forward-facing camera alongside an in-cabin camera. Additionally, configuring a stereo camera setup with two cameras allows for precise distance estimations to objects, further improving AI recognition accuracy in cameras and mobile robots.
Seamless Integration and Design Reusability
Designed with a quad-core Arm® Cortex®-A55 and a single-core Arm Cortex-M33, the RZ/V2N shares its architecture with the RZ/V2H. This makes it easier for developers to reuse software assets from the RZ/V2H. The high compatibility of peripheral functions allows users to select from various RZ/V series MPUs based on AI performance needs, system configurations, and real-time control requirements.
To help users with varying levels of AI expertise, Renesas provides the RZ/V2N AI Software Development Kit. This kit includes evaluation board kits and a software development environment, along with over 50 AI applications for different scenarios. Such resources aim to reduce the time it takes to bring products to market. For those with a background in AI, the kit includes an AI compiler to facilitate the development of personalized AI models, which can be integrated into the RZ/V2N MPUs.
Support from Renesas Ecosystem Partners
Renesas also collaborates with ecosystem partners to offer System on Module (SoM) boards, Single Board Computers (SBCs), and camera modules that incorporate the RZ/V2N. This ecosystem allows developers to focus more on creating AI applications, as the need for hardware design is minimized.
The RZ/V2N, along with its evaluation board (RZ/V2N-EVK), is available for purchase from the Renesas Online Store. For more details, visit www.renesas.com/rzv2n.
* This product is partially based on projects commissioned by the New Energy and Industrial Technology Development Organization (NEDO).
