Renesas RZ/V2H & RZ/V2N: Vision AI MPUs for Embedded Systems

Modern embedded vision systems need to process increasingly complex camera data while keeping power consumption, system size, and latency under control. Renesas' RZ/V family of microprocessors is designed specifically for these types of edge-AI applications.

The RZ/V2H and RZ/V2N are vision AI MPUs that combine Arm processing cores, Renesas' DRP-AI3 accelerator, camera interfaces, image processing, graphics, video codecs, and other peripherals into a single platform.

Both devices are designed for applications such as AI cameras, robotics, industrial automation, machine vision, and intelligent edge systems.

What Is the Renesas RZ/V Series?

The RZ/V family is a series of embedded microprocessors designed for vision AI applications.

Instead of sending camera data to a separate high-power processor or cloud system, the RZ/V architecture enables much of the processing to happen directly at the edge.

A simplified vision pipeline can look like:

Camera Sensor → MIPI CSI-2 → ISP → DRP-AI3 → Computer Vision / AI Application

This approach can reduce latency, lower system power consumption, and enable real-time AI processing directly on an embedded device.

RZ/V2H

The RZ/V2H is the higher-performance device of the two and is designed for demanding vision AI and real-time control applications.

It combines four Arm Cortex-A55 application processor cores running at up to 1.8 GHz with two Cortex-R8 real-time processor cores and a Cortex-M33 sub-core. This heterogeneous architecture allows Linux applications, real-time processing, and lower-level control functions to run on different processing resources.

The RZ/V2H also includes Renesas' DRP-AI3 accelerator. Renesas specifies up to 10 TOPS/W and up to 80 TOPS of AI inference performance using its sparse-model/pruning technology.

This combination makes the RZ/V2H particularly suited to applications where AI processing and real-time control need to operate together.

RZ/V2N

The RZ/V2N is positioned as a more compact, power-efficient vision AI MPU for higher-volume applications.

It uses four Arm Cortex-A55 cores running at up to 1.8 GHz and includes a Cortex-M33 sub-core. Like the RZ/V2H, it incorporates the DRP-AI3 accelerator with approximately 10 TOPS/W power efficiency. Renesas specifies up to 15 TOPS of AI inference performance for the RZ/V2N.

One of the key features of the RZ/V2N is its support for two camera connections, making it well suited to AI cameras, robotics, inspection systems, and other multi-camera applications.

DRP-AI3: AI Processing at the Edge

A major feature of both processors is Renesas' Dynamically Reconfigurable Processor AI (DRP-AI3).

The accelerator is designed to execute AI inference efficiently while reducing the power required compared with relying primarily on a general-purpose CPU or GPU.

AI models developed using frameworks such as ONNX, PyTorch, or TensorFlow can be converted for DRP-AI processing using Renesas' DRP-AI TVM software environment.

This makes the platform useful for applications such as:

Object detection
Object classification
Person detection
Image inspection
Robotics
Machine vision
Smart cameras
Edge AI gateways

The ability to perform inference locally is particularly useful when an application requires low latency or cannot rely on a continuous cloud connection.

Camera and ISP Capabilities

Camera processing is an important part of both platforms.

Both the RZ/V2H and RZ/V2N can use an integrated Arm Mali-C55 ISP, providing hardware image processing for camera applications. The ISP supports resolutions up to 4K.

The ISP provides functions such as:

Auto exposure
Auto white balance
Black-level correction
Demosaicing
Tone mapping
Wide dynamic range correction
Shading correction
Sharpness processing
Noise reduction
Color correction
HDR processing
Image resizing and cropping

These functions allow raw image-sensor data to be converted into processed image data before it reaches the application or AI pipeline.

Camera Interfaces

The two processors differ significantly in their camera connectivity.

The RZ/V2H provides four 4-lane MIPI CSI-2 interfaces, allowing it to support multiple camera streams. The RZ/V2N provides two 4-lane MIPI CSI-2 interfaces.

This makes the RZ/V2H particularly attractive for systems requiring several simultaneous cameras, while the RZ/V2N can provide a smaller and more power-efficient solution for one- or two-camera applications.

A typical camera architecture can be:

Image Sensor → MIPI CSI-2 → Mali-C55 ISP → Image Processing → DRP-AI3 → AI Application

The actual architecture depends on the sensor, camera driver, software framework, and application requirements.

RZ/V2H for Multi-Camera and Real-Time Systems

The RZ/V2H is designed for systems that need both high AI performance and real-time processing.

Its combination of Cortex-A55 application cores, Cortex-R8 real-time cores, Cortex-M33, DRP-AI3, and the additional DRP processing engine makes it suitable for applications where multiple workloads need to operate simultaneously.

Potential applications include:

Industrial robots
Autonomous robots
Machine vision
Factory automation
Multi-camera inspection
Intelligent transportation
Advanced vision systems

The processor also provides interfaces such as PCIe, USB 3.2, Gigabit Ethernet, and CAN-FD for connecting the vision system to other hardware.

RZ/V2N for Compact Vision AI

The RZ/V2N targets applications where a smaller and more power-efficient platform is desirable without giving up dedicated AI acceleration.

Its two MIPI CSI-2 camera interfaces make it well suited for stereo or dual-camera systems, while the DRP-AI3 accelerator provides dedicated AI processing.

Applications can include:

AI cameras
Retail analytics
Smart-city cameras
Industrial inspection
Mobile robots
Vision AI gateways
Intelligent equipment

Renesas specifically positions the RZ/V2N for high-volume vision AI applications where system size, cost, and power efficiency are important considerations.

ISP Tuning and Image Quality

The processor and AI accelerator are only part of an embedded camera system. Image quality depends heavily on the interaction between the sensor, lens, ISP, and tuning parameters.

A typical development process may look like:

Camera Sensor → Sensor Driver → MIPI CSI-2 → Mali-C55 ISP → ISP Tuning → AI Preprocessing → DRP-AI3 → AI Model

ISP tuning can involve exposure, white balance, color correction, noise reduction, HDR, tone mapping, sharpening, and lens-related corrections.

Good image quality can also improve AI performance. Poor exposure, excessive noise, incorrect color, or insufficient dynamic range can make it more difficult for a computer-vision model to identify objects reliably.

RZ/V2H vs. RZ/V2N

The choice between the two platforms depends largely on system requirements.

The RZ/V2H is the stronger choice when a design requires higher AI performance, multiple cameras, and dedicated real-time processing. It provides four MIPI CSI-2 camera interfaces and substantially higher AI performance.

The RZ/V2N is better suited to more compact systems that need strong edge-AI performance with two camera interfaces and lower system complexity.

Both platforms share important technologies, including Cortex-A55 processing, DRP-AI3 acceleration, Mali-C55 ISP support, MIPI CSI-2 camera connectivity, H.264/H.265 video processing, and Mali-G31 graphics.

Embedded Linux and Camera Integration

The RZ/V2H and RZ/V2N are more than AI accelerators. They are complete embedded processing platforms that need to be integrated with an operating system, camera drivers, hardware configuration, and application software.

A camera implementation may involve:

Sensor → Device Tree → Sensor Driver → MIPI CSI-2 → ISP → V4L2 / Camera Framework → AI Application

Developers need to configure the hardware correctly, integrate sensor drivers, establish the camera pipeline, configure the ISP, and optimize data movement between camera processing and AI inference.

This is especially important in multi-camera systems, where bandwidth, memory, synchronization, and processing resources must all be considered.

Conclusion

The Renesas RZ/V2H and RZ/V2N provide powerful platforms for embedded vision and edge AI.

The RZ/V2H targets higher-performance applications requiring multiple cameras, high AI performance, and real-time processing. The RZ/V2N provides a more compact solution for applications that need efficient AI processing and up to two camera connections.

Both combine Arm processing, Renesas DRP-AI3 acceleration, Mali-C55 image processing, MIPI CSI-2 camera interfaces, and embedded software capabilities into platforms designed for intelligent edge applications.

For camera products built around these platforms, successful development requires more than selecting the MPU. Sensor integration, PCB design, camera drivers, Device Tree, ISP tuning, embedded Linux, AI acceleration, and computer-vision software all need to work together.

Building a production-ready vision system around the Renesas RZ/V2H or RZ/V2N requires careful integration of the camera sensors, MIPI interfaces, Mali-C55 ISP, AI accelerator, and embedded software. Explore our hardware design services to learn how we support custom camera integration, ISP development and tuning, schematic and PCB design, and complete embedded vision hardware development for Renesas platforms.