Zest_Sensor_Camera ai

A basic example to demonstrate how to use the Zest_Sensor_Camera board with a TLM CNN wich can recognise handwritten digits.

Requirements

Hardware requirements

The following boards are required:

• Zest Core STM32L4A6RG
• Zest Sensor Camera
• Zest Battery LiPo (optional)

Software requirements

This demo makes use of the following libraries:

• Zest Sensor Camera (include the lm3405 led flash and ov5640 sensor drivers)
• TensorFlow Light for Microcontrollers

Usage

To clone and deploy the project in one command, use mbed import and skip to the target and toolchain definition:

mbed import https://gitlab.com/eg-julien/zest-sensor-camera-ai.git zest-sensor-camera-ai

Alternatively:

• Clone to "zest-sensor-camera-ai" and enter it:

  git clone https://gitlab.com/eg-julien/zest-sensor-camera-ai.git zest-sensor-camera-ai
  cd zest-sensor-camera-ai

• Create an empty Mbed CLI configuration file:

  • On Linux/macOS:

    touch .mbed

  • Or on Windows:

    echo.> .mbed

• Deploy software requirements with:

  mbed deploy

Define your target (eg. ZEST_CORE_STM32L4A6RG) and toolchain:

mbed target ZEST_CORE_STM32L4A6RG
mbed toolchain GCC_ARM

Export to Eclipse IDE with:

mbed export -i eclipse_6tron

Working from command line

Compile the project:

mbed compile

Program the target device (eg. STM32L4A6RG for the Zest_Core_STM32L4A6RG) with a J-Link debug probe:

python dist/program.py STM32L496RG BUILD/ZEST_CORE_STM32L4A6RG/GCC_ARM/zest-sensor-camera-demo.elf

Debug on the target device (eg. STM32L4A6RG for the Zest_Core_STM32L4A6RG) with a J-Link debug probe.

• First, start the GDB server:

  JLinkGDBServer -device STM32L4A6RG

• Then, in another terminal, start the debugger:

  arm-none-eabi-gdb BUILD/ZEST_CORE_STM32L4A6RG/GCC_ARM/zest-sensor-camera-demo.elf

Note: You may have to adjust your GDB auto-loading safe path or disable it completely by adding a .gdbinit file in your $HOME folder containing:

set autoload safe-path /