Avnet has released a new single-board computer with a unique CPU designed exclusively for visual AI development. The New RZBoard V2L provides energy-efficient vision AI.
Avnet is having a week of new product releases. The RZBoard V2L was announced today, after the announcement of the ZUBoard 1CG Development Kit earlier this week. The RZBoard V2L is a single-board computer in the Raspberry Pi form size with a custom CPU for vision-artificial intelligence (AI) research.
The RZBoard is a cost-effective platform for visual AI development, with a price tag of $149. The standard 40-pin header interface on the Raspberry Pi is also meant to allow users to use their existing collection of header peripheral boards and speed up prototyping.
With the following lengthy list of peripherals, the RZBoard covers almost all of the interfaces required for visual AI rapid development:
- Ethernet Gigabit
- Wi-Fi 802.11ac and Bluetooth 5
- There are two USB 2.0 host ports.
- OTG USB 2.0 interface
- Camera Serial Interface (CSI) and MIPI Display Serial Interface (DSI) (CSI)
- Flexible Data-Rate Interface for Controller Area Networks
- 40-pin extension header compatible with Pi-Hat
- The header for shuttle expansion
- For interacting with analog sensors, there are five 12-bit analog-to-digital converter inputs.
- USB-C connection provides 5V input power.
- For voice command and audio playback, there's an audio codec and a stereo headphone port.
A general-purpose microprocessor with a 1.2GHz dual-core Arm Cortex-A55 and a 200MHz ARM Cortex-M33, as well as Renesas' AI accelerator "DRP-AI" for vision, lies at the heart of the RZBoard V2L. The built-in 3D graphics engine with Arm Mali-G31, video codec (H.264), and 16-bit DDR3L/DDR4 interface help this chip handle its video workload.
Renesas' DRP-AI AI accelerator is based on the company's Dynamically Reconfigurable Processor (DRP) special-purpose hardware. The DRP-power AI's efficiency does not require external heat dissipation such as heat sinks or cooling fans, which is crucial to the RzBoard's power efficiency.
The DRP-AI also incorporates real-time AI inference and image processing operations, including color correction and noise reduction, which eliminates the need for an external image signal processor (ISP).
With two types of processors, the DRP-AI handles neural network layers: the Dynamically Reconfigurable Processor (DRP) for pooling and other layers, and the multiply-accumulate processor (AI-MAC) for convolution and fully connected layers.
DRP-AI Translator, a software utility for the host PC that converts a trained AI model to DRP-AI object code, appears to be only available for Ubuntu 18.04 at the time. However, it looks that the software package for creating code for Arm cores is accessible for both Windows and Linux.
The RzBoard V2L is now available for pre-order, with delivery expected in Q4 2022. A 5V/3A USB Type C power supply, MIPI CSI camera, and a 7-inch display will be offered as accessories for the RzBoard. It will also be on exhibit at the Embedded Vision Summit in Santa Clara, California on May 17-19, 2022!
Posts
0 Comments