Click here to see the project on EdgeImpulse.
A deep learning project that uses Edge Impulse to train a model to learn equating specific movement to a specific output.
Note: Make sure you have the
edgeImpulse CLIinstalled. You can find installation instructions here.This project uses the
edge-impulse-data-forwardertool.
| component name | how many | link |
|---|---|---|
| Arduino Nano 33 BLE | 1 | |
| 47.5k Resistor | 3 | |
| Flex sensors | 3 | sparkFun |
| PCB Terminal Block | 3 | RS Pro |
I used this hookup guide by SparkFun to setup my system.
Note: They wired it to the 5v line, but one issue I had was 5v was not being output by my Arduino. This is likely because I was powering my Arduino from my laptop. 3.3v works fine with pull-up resistors, rather than pull-down resisors.
It looks like this:
My wiring diagram is below.
Open the arduino script, and push the code to the device. I opened my serial port to verify that the values were being output correctly.
I opened my terminal, and ran edge-impulse-data-forwarder. Then closed my serial port in Arduino IDE.
.
When first running the data forwarder, be aware you must create the device and link to it in the edgeImpulse website.
By clicking on the Edge Impulse link above, it will take you to the pre-made project. If you click Model Testing in the left bar, you can run tests with a connected device. Otherwise clicking on the Deployment option allows you to use this model in your projects.
I've found the best way to train the data is to set the sample time to a higher value like 10 seconds. Slowly complete the movement desired. Then clicking the 3 dots beside the dataset, you can split the sample. From here you can rapidly create several data samples for the model to read.