Coral Dev Board Micro
Linux or macOS computer with Python 3 installed
USB-C cable (to connect the board to your computer)
Although not required for this setup, you should also get a set of header pins that can be soldered onto the board and a USB-to-TTL serial cable (such as this one by Adafruit) so you can connect to the board's serial console.
 
The hardware
Aside from the Coral Edge TPU coprocessor for accelerating TensorFlow Lite models, the Dev Board Micro includes the following processor and memory:
Dual-core i.MX RT1176 MCU
Cortex-M7 (@800 MHz) and a Cortex-M4 (@400 MHz)
2 MB of built-in SRAM
64 MB of SDRAM
More than 1 GB of flash memory
 
Color camera (324 x 324 px)
PDM microphone
2 buttons:
User programmable button (also toggles power with a 7-second hold)
MCU reset button
4 LEDs:
Camera LED (indicates camera operation)
Edge TPU LED (indicates Edge TPU operation)
User LED (application behaviors)
Status LED (indicates board status)
2 GPIO headers with 12 pins (digital, analog, and power pins)
Board-to-board connectors for add-on boards such as the Coral Wireless Add-on and PoE Add-on
For more hardware details, see the Dev Board Micro datasheet.
The software
Because the main processor is a microcontroller (MCU), the Dev Board Micro runs small programs without a general-purpose operating system (there's no Linux OS like on other Coral boards). The platform we've built for Dev Board Micro is based on FreeRTOS and we also offer compatibility with Arduino. (Alternatively, you can write bare-metal apps with the MCUXpresso SDK, but our documentation covers only FreeRTOS and Arduino.)
All apps are written in C/C++ and the coralmicro API library provides a variety of APIs to help you build apps for the Dev Board Micro. For example, it includes APIs to use the on-board camera, microphone, and GPIO pins, plus APIs to read/write files, pass messages between the two MCU cores, connect to Wi-Fi (requires the Coral Wireless Add-on board), and much more.
As for running ML models, you can execute TensorFlow Lite models either on the MCU or with acceleration on the Edge TPU, using TensorFlow Lite for Microcontrollers (TFLM). The coralmicro library also provides APIs to simplify a lot of this code for you.
Now let's run some ML models on the board.