Repolode
Submit
jomjol

jomjol/AI-on-the-edge-device

An ESP32-CAM based device that photographs analog water, gas or power meters and uses on-device AI to convert the readings into digital values.

View on GitHubWebsite
Stars
8,884
Forks
929
Open issues
205
Last commit
yesterday

What is it?

AI on the Edge Device connects existing non-digital meters, such as water, gas and electricity meters, to digital systems without replacing them. An ESP32 board with a supported camera captures photos of the meter at set intervals, extracts the regions of interest from the image, and runs them through a TensorFlow Lite model to read the digits.

The firmware handles inline image processing such as feature detection, alignment and ROI extraction, and includes an integrated camera and illumination. Readings can be sent to an MQTT broker, written to InfluxDB 1 or 2, or exposed through a REST API, and the device integrates with Home Assistant. A web interface handles administration, and firmware can be updated over the air.

The hardware is a small, low-cost board, and 3D-printable housings for water, power and gas meters are published alongside the project. Setup is documented on a separate MkDocs site, with flashing done through a browser-based web installer or standard ESP32 tools.

Key features

  • 01Recognizes meter digits with an on-device TensorFlow Lite model.
  • 02Performs inline image processing for alignment and ROI extraction.
  • 03Publishes readings via MQTT, InfluxDB 1/2, or a REST API.
  • 04Integrates with Home Assistant.
  • 05Provides a web interface for administration and supports over-the-air updates.

Use cases

  • Logging water or gas consumption from an existing analog meter into a home dashboard.
  • Feeding electricity meter readings into Home Assistant or InfluxDB without replacing the meter.
  • Building a low-cost smart-meter setup on an ESP32-CAM.

Quick start

  1. 1

    Download the latest release from the Releases page

  2. 2

    Flash the ESP32 over USB with the browser-based Web Installer

  3. 3

    Set up the SD card automatically after firmware install, or manually following the documentation

Things to know

  • Requires a supported ESP32 camera board and a physical analog meter to photograph.
  • The license is not clearly identified in the project metadata.

In collections

Computer vision libraries & toolsEmbedded systems & microcontroller projects

Topics

#esp32#edge-ai#tensorflow-lite#iot#home-assistant#mqtt#computer-vision#embedded

FAQ

What is AI-on-the-edge-device?
An ESP32-CAM based device that photographs analog water, gas or power meters and uses on-device AI to convert the readings into digital values. AI on the Edge Device connects existing non-digital meters, such as water, gas and electricity meters, to digital systems without replacing them. An ESP32 board with a supported camera captures photos of the meter at set intervals, extracts the regions of interest from the image, and runs them through a TensorFlow Lite model to read the digits.
Is AI-on-the-edge-device free to use? What license does it have?
AI-on-the-edge-device is open source under the Other license. Check the license terms before using it in commercial products.
Is AI-on-the-edge-device still maintained?
Actively maintained. The last commit to AI-on-the-edge-device was on Oct 23, 2026.
What can I use AI-on-the-edge-device for?
Logging water or gas consumption from an existing analog meter into a home dashboard. Feeding electricity meter readings into Home Assistant or InfluxDB without replacing the meter. Building a low-cost smart-meter setup on an ESP32-CAM.
How do I install AI-on-the-edge-device?
Download the latest release from the Releases page Flash the ESP32 over USB with the browser-based Web Installer Set up the SD card automatically after firmware install, or manually following the documentation
Where can I get AI-on-the-edge-device?
The source code and installation instructions are on GitHub: https://github.com/jomjol/AI-on-the-edge-device

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects