Edge Impulse Example: stand-alone inferencing (Arducam Pico4ML)

This builds an exported impulse to a .uf2 file that can be dragged and dropped to Arducam Pico4ML.

Test use

Connect the Arducam Pico4ML to the PC using the micro-USB cable and make sure to force it to USB Mass Storage Mode by pressing and holding the BOOTSEL button. A window named RPI-RP2 will appear on the PC, just drag the bin/magic_wand.uf2 file to the RPI-RP2 window!

Recompile

1. Decompress the files

Unzip the C++ library generated by the Edge Impulse project and copy and overwrite it to the current directory

CMakeLists.txt file without copying

.
├─ bin
├─ CMakeLists.txt
├─ edge-impulse-sdk
├─ ICM20948
├─ LICENSE
├─ model-parameters
├─ pico_sdk_import.cmake
├─ README.md
├─ source
├─ tflite-model
└─ tool

2. Compile

1. Create the build folder

   mkdir build && cd build

2. Compile

   camke .. && cmake --build . --parallel

3. Write the .uf2 file

Connect the Arducam Pico4ML to the PC using the micro-USB cable and make sure to force it to USB Mass Storage Mode by pressing and holding the BOOTSEL button. A window named RPI-RP2 will appear on the PC, just drag the build/magic_wand.uf2 file to the RPI-RP2 window!

Data conversion

1. go to the tool folder

2. Install the python dependency library

   pip install -r requirements.txt

3. Convert the data

   Convert the json file downloaded in the previous step to a png file

   python json2png.py --dataset wanddata.json
   

   Under the datasets folder you can see the converted png file