Upgrade from camera image recognition to measuring vibrations from plant roots

So far we don't have access to sensor data. If we had some, we could retrain the model to recognize signals from that (rather than from the camera).

Downside: this would require the end user to have expensive sensor equipment to use the app.

The issue names no files, tests, or entry points; first identify where the current camera-based recognition model and app flow are implemented. Confirm whether suitable vibration sensor data exists and define the retraining and sensor-equipment requirements before implementation. Done would require an agreed sensor data source and a tested replacement for camera-based recognition.

Written by the indexing model from the issue text.