When an IoT device needs AI-based vision, the trade-off between processing capability and power consumption can be tricky — especially for battery-based designs. Also, where there is high power consumption, there is heat which might mean bulky heat sinking. Fortunately, Canada-based engineering design group MistyWest partnered with Renesas to address these challenges. The result is a battery-powered system-on-a-module for AI vision applications called MistySOM.
In its 40 by 50 millimeter form factor, MistySOM includes:
- A Renesas RZ/V2L microprocessor
- 2 GB of DDR4 RAM
- 32 GB of eMMC storage
- An extensive array of communication interfaces
The 123 configurable GPIOs support UART, I2C, Octal SPI Flash, USB 2.0, CAN FD, SDXC, MIPI CSI/DSI, Gigabit ethernet, a 12-bit ADC, and SSI audio — just to name a few. Check this MistySOM product brief for a complete list of available peripherals.
The Renesas RZ/V2L is a dual-core Arm Cortex-A55 CPU running at 1.2 GHz. The RZ/V2L also features Dynamically Reconfigurable Processor (DRP) Technology. The DRP-AI accelerator allows a vision algorithm to change configurations on-the-fly.
For example, you can break down a complex algorithm into smaller contexts. Then those can be multiplexed in the data path as needed. The configuration change can happen as fast as one millisecond. Two real-world applications for DRP-AI are 2D barcode scanning and iris detection.
This integrated combination allows MistySOM to provide a real-time AI inference engine with hardware-accelerated image processing functions, like noise reduction.
MistyWest plans to sell the module through a GroupGets campaign. After the SOM launches, a carrier board will be available in the future. Visit the MistySOM campaign page to sign-up for notifications when the campaign goes live.
For those attending embedded Vision Summit 2022 in Santa Clara, you can meet with MistyWest at the Renesas booth.