Awesome

Magic Wand with Machine Learning

Magic Wand using ESPectro32 board, powered by TensorFlow Lite for Microcontrollers and PlatformIO. Previously I made a similar project but using Arduino Nano 33 BLE Sense board.

As ESPectro32 is based on ESP32, this project should be easily adapted for other ESP32 board. You can change this to #define ESPECTRO32 0 in output_handler.cpp. But you need to have some kind of display to let user know what is the predicted gesture. ESPectro32 has on-board LED Matrix to display the predicted gesture.

Demo

Magic wand on ESPectro32 board

Interaction demo video is quite similar to this:

Prerequisites

• PlatformIO
• PlatformIO's platform-espressif32. Should be installed automatically

Train

You can train the Machine Learning model to recognize your own gesture, or even other gestures than Magic Wand. The model and training code is here.

I'm thinking to create some guideline to do that training in Azure Machine Learning, let me find the time.

Credit

• Magic Wand original sample code
• Some code is inspired from here
• SparkFun_LSM6DS3_Arduino_Library. Need to include here, as its PlatformIO library is failed to compile.