The Google Coral Dev Board Micro is a microcontroller board with a dual-core MCU, Coral Edge TPU, a Himax HM01B0 CMOS color camera (324 x 324 px or QVGA (320 x 240 px); 110º diagonal FOV; ƒ/2.0 focal ratio; Fixed focus), a PDM microphone, user and reset buttons, four LEDs and 2 GPIO headers with 12 pins (digital, analog, and power pins).
With this board, you can build low-power systems with fast on-device inferencing for vision and audio ML applications. You can also expand the hardware with custom add-on boards using the high-density board-to-board connectors.
The firmware development platform is based on FreeRTOS with a CMake toolchain. It also includes support for development with Arduino. The FreeRTOS plateform for coralmicro includes APIs to use the Dev Board Micro's camera, microphone, and GPIOs, plus APIs for reading/writing files, creating RPC services, communicating across MCU cores, connecting to Wi-Fi (requires the Wireless Add-on board), and TensorFlow Lite for Microcontrollers to execute ML and DL models on either the MCU or with acceleration on the Edge TPU. Bare-metal apps development with the MCUXpresso SDK is available but it is not documented by Google.
The power consumed by the Dev Board Micro depends on a variety of application behaviors, but most importantly, the extent to which the Edge TPU is being used. Typical test scenarios when the Edge TPU is active show average power peaks around 3 W. Beware that, depending on the type of ML model you're using, the Edge TPU can cause significant spikes in current draw.
| Name | Color | Description |
|---|---|---|
| Camera | Green | Indicates camera activity. By default, it is on during image capture, for a minimum of 500 ms. |
| TPU | White | Indicates Edge TPU status. By default, it is on while the Edge TPU is powered. It is programmable, but only while the Edge TPU is powered. |
| User | Green | Programmable. Off by default. |
| Status | Orange | Indicates board operating status. By default, it turns on briefly when the board boots from flash and it is programmable. |
https://coral.ai/docs/dev-board-micro/get-started/#1-gather-requirements
https://github.com/google-coral/coralmicro
https://coral.ai/docs/dev-board-micro/arduino/#1-gather-requirements
Select the following examples in section called "Examples for Dev Board Micro" in the examples menu
File > Examples > TensorFlow > DetectFaces.File > Examples > TensorFlow > DetectObjects.
- Classify Images with
mobilenet_v1_1.0_224_quant_edgetpu.tflite(labels) - Classify Speech
See also Pretrained models.
UART6_TXandUART6_RXare configured for serial port communication and should not be used for other purposes.
Connect the LoRaWAN E5 modem to the UART pins (UART6_TX and UART6_RX).
TODO
TODO
TODO
Connect a Bosch Sensortec BME280 breakout to the I2C pins
Install the Arduino lib https://www.arduinolibraries.info/libraries/bme280
The original driver is https://github.com/boschsensortec/BME280_driver
TODO
TODO
https://coral.ai/docs/dev-board-micro/wireless-datasheet/