Is it possible to do the gesture recognition using raspberry pi 4 by using connecting it with adxl 345 accelerometer or by connecting it with Arduino 33 ble sense and sending the accelerometer data from it
Can you please describe your use case further? You are able to do all gesture recognition and inferencing directly on the Arduino Nano 33 itself without having to send all of the data to a Raspberry Pi. Check out the docs here: https://docs.edgeimpulse.com/docs/development-boards/arduino-nano-33-ble-sense & Continuous motion recognition - Edge Impulse Documentation
I want to run a gesture recognition model in raspberry pi using gesture data either from adxl 345 accelerometer or by passing accelerometer data from nano 33 ble to raspberry pi.
I need it to be done using the raspberry pi, since I am using that result for further application.
It will be very helpfull if I am able to classify the gesture within the raspberry pi itself
( I have implemented the gesture recognition in Nano 33 ble separately )
Then in this case you will need to write an application that sends the raw accelerometer data from the Arduino Nano 33 BLE over some connection like USB Serial/Wifi/BLE to the Raspberry Pi. Then on the Pi, you can have an application that polls for data on that connection and retrieves the inferencing result from your trained model that has already been deployed on the Pi. Check out the following documentation:
- Get the acceleromter data from the Arduino Nano 33 BLE Sense into Edge Impulse using the Edge Impulse CLI or Data Forwarder
- Design and train your gesture recognition model using that accelerometer data: Continuous motion recognition - Edge Impulse Documentation
- Deploy this model to your Raspberry Pi via the Edge Impulse Linux Runner or Python SDK, Raspberry Pi 4 - Edge Impulse Documentation etc.
- Follow examples from the guides below to get new data off of your Arduino Nano 33 BLE Sense and onto the Raspberry Pi
- Pass that retrieved data into the Raspberry Pi’s deployed Edge Impulse model and collect inference result
Can we do live classification like this ?
From what I understood from the custom classification script is that , the data should be in a text file format ( is it right? ) .
linux-sdk-python/classify.py at master · edgeimpulse/linux-sdk-python · GitHub
And if so is there any particular format in which the data should be represented in the text file