Now that you have this working...
-
Load the code to record gestures. Create additional CSV files with more gestures. Retrain the model in Colab. Load the new model back onto the Arduino.
- Note: you'll need to edit the code to add the names of the new gesture files
-
Try increasing and decreasing the number of recordings per gesture, how does this impact performance?
-
Try to only use the accelerometer or gyroscope data (not both), how does this impact performance?
-
Tweak the model structure and parameters
- Can you get better results?
- Can you reduce the size and still get "good" results
-
Grab a board and the TinyML book to continue at home.