Listen to the meter

The Meter
Whisperer

Small leaks and abnormal water use rarely announce themselves. They persist in the background — hidden behind walls, inside tanks, or in the meter's own quiet rotation.

04 / The signal

Raw vibration
becomes legible.

One time window. Four ways to see the same invisible event.

Simulated sensor window · 8.0s
Interactive visual
raw acceleration
0s time window 8s

How it works

The meter already knows.

Water movement causes mechanical activity inside a traditional meter. That activity produces vibration. A clip-on accelerometer turns those vibrations into a signal, and a trained model turns that signal into a calm, actionable explanation.

1CaptureClip-on accelerometer
2ExtractTime + frequency features
3ClassifyRandom Forest model
4ExplainSeverity + action

Interactive AI demonstration

Hear the leak.

Change the pattern to see how the product turns signal into a calm, actionable explanation.

Interactive AI demonstration
Activity pattern
Waiting for backend…
AI interpretation Inference
No abnormal activity
Severity: — Confidence: — RMS: —

Real sensor recordings

Load a CSV from data/raw/ and run a real prediction.

Live leak check

Try it with your own recording.

Upload a .csv recorder file in the same format as the existing recordings. The model will classify it, identify the probable leak type, explain the reason, and give suggestions for improvement.

Live preview
Sensor file

Choose a CSV file to run the live preview.

Same columns as the data/raw/ recordings: time_s, acc_x_m_s2, acc_y_m_s2, acc_z_m_s2.