A microcontroller board containing a camera, microphone, and Coral Edge TPU.

NXP iMX RT1176 #Arm Cortex-M7/M4 crossover processor and Coral Edge module power the upcoming Google Coral Dev Board Micro in RaspberryPi Zero physical factor with camera and microphone.

Google Unveils the Coral Dev Board Micro, Its First TinyML Edge AI Board Based on a Microcontroller

The next addition to Google's Coral line of low-power edge AI development boards is also the company's first microcontroller board, which will be available "soon."

The Coral Dev Board Micro is the latest addition to Google's Coral line of low-power edge AI development boards, this time featuring the Tensor Processing Unit (TPU) on a breadboard-friendly microcontroller alongside Arm Cortex-M4 and Cortex-M7 processors.

StackyPi Supports Raspberry Pi HAT's Available at £7 on Kickstarter

Three years ago, Google introduced the Coral line, which included a Raspberry Pi-like development board and a USB accelerator, both of which were intended to demonstrate the low-power, high-performance capabilities of its own Tensor Processing Unit (TPU) accelerator technology. The Coral DevBoard Mini was introduced to the line-up last year, and now the business is back with an even more affordable gadget – its first to focus on microcontroller-powered TinyML applications.

Google describes its new board as "a microcontroller board with a built-in camera, microphone, and Coral Edge TPU, letting you to quickly prototype and deploy low-power embedded systems with on-device ML inferencing." "You can develop systems that seamlessly cascade from extreme low-power ML inferencing to more complicated — yet still power-efficient — ML inferencing by integrating the Cortex M4 and M7 CPUs with the Coral Edge TPU on this board."

The NXP i.MX RT1176 system-on-chip, which comprises a single Arm Cortex-M7 core and a single Arm Cortex-M4 core, is at the heart of the Coral Dev Board Micro. Although clock rates have not been disclosed, comparable implementations clock the Cortex-M7 at up to 1GHz and the Cortex-M4 at up to 400MHz, putting them in the highest echelon of microcontroller performance.

Google has added an Edge TPU coprocessor to this, which offers four trillion operations per second (TOPS) of INT8 computation capability in a 2W power envelope – measured only for the coprocessor and excluding the i.MX or other components. On-board memory is 64MB, with 128MB of flash memory.

A low-resolution 324 324 color camera sensor and a digital PDM microphone are also included on the board, as well as two 12-pin general-purpose input/output (GPIO) headers on either side of the board that are unpopulated by default and oriented to be breadboard-compatible. The board's extra pins are carried out to a pair of high-density 100-pin board-to-board connections for those who require more. Finally, there are two buttons and four LEDs.

There is no network connectivity, which is odd considering the company's description of the device as a microcontroller board for embedded projects. Power and data are provided via a single USB Type-C connector, with a Wi-Fi and Bluetooth module available as an add-on for customers who want wireless connectivity. A carrier board will provide wired Ethernet connectivity as well as support for Power-over-Ethernet (PoE).

However, there are two parts of the board that Google isn't ready to reveal just yet: Pricing and availability are available. The device's official product listing just mentions that it is "coming soon," implying that interested parties will have to wait a while before getting their hands on it.

Posts You May like: 

Read more »