Stopping extreme water consumption with tinyML

Because the frequency and depth of droughts around the globe continues to extend, with the ability to scale back our water utilization is significant for sustaining already strained freshwater assets. And in response to the EPA, leaving a faucet operating, whether or not deliberately or by chance for simply 5 minutes can devour over ten gallons of water. Nevertheless, Naveen has leveraged the ability of machine studying to construct a tool that may routinely detect operating taps and ship alerts over a mobile community in response.

The {hardware} for this undertaking is primarily centered round a Blues Wi-fi Notecard for mobile connectivity, a Blues Wi-fi Notecarrier-B as its breakout board, and a machine learning-capable microcontroller within the type of an Arduino Nano 33 BLE Sense. Past merely having a 32-bit Arm Cortex-M4 processor and 1MB of flash storage, its built-in microphone can be utilized to simply seize audio information. On this undertaking, Naveen uploaded a dataset containing quarter-hour of both faucet noises or background noise into the Edge Impulse Studio earlier than coaching a 1D convolutional neural community, which achieved an accuracy of 99.2%.

From right here, a brand new Twilio route was created that permits the Blues Wi-fi Notecard to generate SMS messages by sending an API request. Now each time a faucet has been categorized as operating for too lengthy, the Nano 33 BLE Sense can transmit a easy command over I2C to the Notecard and alert the recipient.

For extra details about this undertaking, you possibly can learn Naveen’s write-up right here on Hackster.io.