compareeHub

A CAMERA TAPED TO YOUR WATER METER READS IT FOR YOU, NO METER SWAP NEEDED

An ESP32-CAM board taped to your water meter reads the dials with a neural network running on the chip itself, no cloud, no meter swap.

by jomjol

FULL CAD BOM FIRMWARE DOCS

HomeAutomation

Built withESP323D printing

difficulty
●●●○○
time
a weekend
license
Custom Dual Use License
repo
repo ACTIVE8,832 stars

WHAT YOU’LL NEED

Videos
1
Jump to section

COMPAREE VERDICT

AI-on-the-edge-device does exactly what it says: it points a camera at the dial, crops the regions where the digits sit, and runs TensorFlow Lite on an ESP32-CAM to read them. The whole device is around 3 x 4.5 x 2 cm, costs less than ten euros, and works on water, gas and electricity meters without touching the meter itself. That last part is the real win: just a printed holder and a camera. The firmware has a web interface for configuration, OTA updates, Home Assistant integration, MQTT, InfluxDB and a REST API. The repository is actively developed, well documented and has over 8,800 stars. The catch: the licence is a custom Dual Use License that forbids commercial use without a separate licence, so this is for private installs. What eats time is buying a board with enough PSRAM, choosing the right model and dialling in the ROI boxes. If you already run Home Assistant and have flashed firmware before, this is a weekend well spent.

GOOD TO KNOW

  • —3D-printable housings for water, power and gas meters are linked from the README and hosted on Thingiverse.
  • —No formal bill of materials: the hardware is an ESP32-CAM with a supported camera, an SD card and a 5V supply, and the docs' Hardware Compatibility page lists tested boards, PSRAM chips and SD cards.
  • —Firmware is pre-compiled and flashable over USB; source is available for custom builds.
  • —Documentation covers installation, web interface configuration, ROI setup and integration with Home Assistant, MQTT and InfluxDB.
  • —The licence is a custom Dual Use License, not a standard open-source licence, and restricts commercial use without prior approval.
  • —No official video walkthrough exists; the good setup guides are community-made.

Parts to buy

3 items

From our check of the build. Exact quantities and part numbers are in the creator’s BOM.

  • MicroSD cardFind
  • USB-to-serial adapter for the first flashFind
  • 5V power supplyFind

BUILDS OF THE WEEK

Five open-source builds worth your weekend, every week.

Checked like this one: what’s really in the repo, what it costs, how hard it is. One email, unsubscribe anytime.

Can I build this?

PrintOne housing from the designs linked in the README's Casing section (water, power or gas meter, camera-only housing, or top cover with antenna holder), matched to your meter.
BuyAn ESP32-CAM board with OV2640 camera and at least 4 MB of PSRAM (some boards are mislabelled), a microSD card (the docs note that smaller cards up to 4 GB tend to be more stable), a USB-to-serial adapter for the first flash, and a 5V power supply. The board's built-in LED provides the illumination.
ToolsA computer with a browser for the web installer (or esptool), a USB-to-serial adapter, a screwdriver for mounting, and Home Assistant, an MQTT broker or InfluxDB to receive the readings.
SkillsFlashing ESP32 firmware over USB, configuring a web interface, setting ROI boxes on a live camera feed, and basic MQTT or Home Assistant integration. The ROI setup is the fiddly part and requires patience.
TimeA weekend if you have flashed ESP32 boards before and your meter is accessible. A weekend-plus if the mounting location is awkward or you are learning the ROI tool for the first time.
Cost$—dominated by the ESP32-CAM board (under ten euros) and the power supply. The SD card and USB adapter are a few euros more.
Safety5V DC throughout, no mains voltage in the device itself. If you are powering it from a mains adapter, standard plug-in power supply rules apply. Do not open or modify the utility meter.

Build at your own risk. Projects involve tools, electronics and sometimes mains voltage — follow the creator’s safety notes.

Videos

Community walkthrough video. Not an official project channel.

More builds like this

All projects
TeddyCloud is an alternative server for your TonieboxHomeYou can build a Wi-Fi motion sensor that works without a cameraHomeWLED runs addressable LED strips from a cheap board with no app, no account and no cloudHome

Gallery

Start here

Navigation into the creator’s own docs — we don’t rewrite the guide, we route you to the source.

  1. 1.Pick the housing for your meter type from the README's Casing section (Meter housings vary widely; picking the wrong STL wastes a full print.)
  2. 2.Flash the firmware to the ESP32-CAM using the web flasher or esptool (The web flasher is the fastest route if you are on a supported browser.)
  3. 3.Mount the device, power it on, and connect to its Wi-Fi access point to configure your network(Initial setup happens over a temporary AP; once configured, it joins your home network.)
  4. 4.Use the web interface to set the ROI boxes around each digit on the live camera feed (This is the step that takes iteration. Start with the default model and refine.)

Resources

Documentation, files and community threads for this build — we link straight to the original sources and never rehost the creator’s files.

KNOWN ISSUES

  • The licence is a custom Dual Use License that forbids commercial use without prior approval. If you are reading meters for anyone other than yourself, check the terms first.
  • ROI setup is fiddly and meter-specific. Budget an hour to get the boxes aligned and the recognition stable, even with good documentation.
  • Not every ESP32-CAM board works. You need at least 4 MB of PSRAM, and some boards are labelled 4 MB but only have 2 MB. Check the Hardware Compatibility page before you buy; it lists tested OV2640 camera modules too.
  • Meters usually sit in dark places, so the device lights the meter with the ESP32-CAM's built-in LED. Reflections on the meter glass can still spoil the image, so check the live picture while you mount it.
  • The device needs constant 5V power and Wi-Fi. If the meter is in a basement with no outlet or weak signal, solve that before you print anything.
  • There is no single official video. The README links a list of community-made YouTube walkthroughs instead.

Does this work on digital meters?

It is built mainly for meters with rotating dials and mechanical digit wheels, but the documentation lists a digit model (dig-class11) that is well suited for LCD digits. If your meter already offers an optical or radio interface, that may be the simpler route.

Can I use this commercially, say to read meters for tenants?

Not without permission. The licence explicitly restricts commercial use. Contact the project maintainer if you need a commercial licence.

How accurate is the recognition?

It depends on your meter, the image quality and the ROI setup. The project ships several models for digits and analogue pointers, and the docs explain which to choose and how the networks are trained.

What happens if the Wi-Fi goes down?

With data logging on (the default), readings are written to CSV files on the SD card, kept for 3 days by default. The docs warn that frequent writes can wear out the SD card.

Do I need to break the utility seal to install this?

No. The device mounts externally and only looks at the meter face. The meter itself is untouched.

Community builds

No community builds yet — be the first, we feature the best ones.

Discussion1

FROM THE COMPAREE TEAM

Over 8,800 stars and the device costs less than ten euros, so the barrier is not the hardware. What stopped you the last time you looked at this: the ROI setup, the mounting location, or something else?

CompareeTEAM2mo agoedited

Practical notes from our verification: the repository, star count and licence terms were checked directly on GitHub, and the project is still actively developed under jomjol's account. The custom Dual Use License is stated in Licence.md in the repository root: private, non-commercial use, modification and sharing are free, while any commercial use needs a separate commercial license from the rights holder, so this is not a standard open-source licence. The README links a list of community-made YouTube walkthroughs rather than one official video. The cost figure (less than 10 EUR) comes from the README and refers to the small ESP32 camera device itself. Printable cases for water, electricity and gas meters are linked from the README's Casing section (on Thingiverse), and the full setup is covered in the official documentation site, so the missing piece is not documentation, it is the patience required to align the ROI boxes on your specific meter. Correction (4 October 2026): we re-checked this page line by line against the project's own repository, documentation and videos, and fixed errors in earlier versions.

jomjol

jomjol (Josef Müller) started AI-on-the-edge-device to show that edge AI can run on a low-cost ESP32 and digitise ordinary water, gas and electricity meters. It is now maintained with a group of contributors.

GitHub

Star the project on GitHub

DISCLAIMER

  • Comparee is not the author of the projects featured here. All rights to each project belong to its creator — every page links to the original source, and we never host creators’ files.
  • Information is provided without warranty and may become outdated as projects evolve. Prices are indicative bands only — always check the creator’s parts list for current costs.
  • Building and operating any project is at your own responsibility. Protective equipment, safe workshop practice and compliance with local regulations are the builder’s responsibility.
GitHub repo