Watch: Community water-meter tutorial (German)

Meine smarte Welt · third-party tutorial · Original project source ↗

Watch: Community guide to common mistakes (German)

Alex Kly · third-party tutorial · Original project source ↗

CAIBI ASSESSMENT

AI-on-the-edge-device

CAIBI VERDICTBUILD IT

An ESP32-CAM project that reads existing water, gas or electricity meters and publishes the readings digitally.

See what the build involves Original source ↗

Build or buy?

Compare the documented build with the purchase reference.

Build it yourself

Build estimate£25–£60CAIBI estimate · low confidenceOne meter retrofit; existing meter and server required

Time: Varies
Difficulty: 3/5
Physical processes: ESP32-CAM

See build requirements →

Purchase reference

Buy price referencePrice not checkedNo priced alternative recorded

Check the source for the exact model, availability and what is included.

Check comparison evidence →

Purchase comparison not yet priced

What the prices cover

A practical allowance for ESP32-CAM hardware, lighting/mounting and small installation materials. The project reuses an existing meter, so installation choices dominate the total.

Includes: ESP32-CAM route; Basic illumination/mounting; Small power and fixing items.

Not included: Existing utility meter; Custom enclosure/service work; Home automation server.

Build evidence recorded 4 Sep 2026. Currency conversions are approximate.

A replacement smart meter or separate energy monitor changes the measurement route rather than doing the same retrofit job, so CAIBI does not claim a like-for-like saving.

BUILD AT A GLANCE

Source view and main parts

Check the source ↗

AI-on-the-edge-device · jomjol ↗

PHYSICAL BUILD6 key parts or groups

No.	Source specification
01	ESP32 board with a supported camera
02	Integrated or external illumination
03	FAT formatted SD card
04	USB connection for initial flashing
05	Meter-specific mounting or 3D printed housing
06	Wi-Fi network for setup

Full source specifications

ESP32 board with a supported camera
Integrated or external illumination
FAT formatted SD card
USB connection for initial flashing
Meter-specific mounting or 3D printed housing
Wi-Fi network for setup, updates and integrations

TOOLS / PROCESSES1 involved

ESP32-CAM

The real-world work

YOU STILL HAVE TOThe real-world work

Mount the camera rigidly in front of the real meter
Create stable lighting and remove glare from the physical installation
Choose and adjust the final camera angle and distance
Verify that the recognised reading matches the actual meter over time

THE BUILD SEQUENCE

How it comes together

1Preparation

Check camera hardware compatibility
Start with an ESP32 camera board supported by the project documentation. Camera choice and illumination directly affect recognition quality.
YOU
Flash the current release
Install the firmware over USB using the project web installer, Espressif tooling or esptool. Later updates can be installed over Wi-Fi.
AI + YOU

2Build and assembly

Prepare the SD card
Use the built-in setup route or prepare the FAT formatted SD card manually as documented.
AI + YOU
Mount and light the meter
Fit the camera so the meter face is stable, well lit and repeatable. This physical setup is part of the recognition system, not a cosmetic detail.
YOU

3Configure and test

Configure the reading regions
Use the web interface to align the image and define the regions that the model should read. AI can help interpret settings and troubleshoot, but you must validate the actual readings.
AI + YOU
Send the readings somewhere useful
Connect the finished device to MQTT, InfluxDB, Home Assistant or the REST API once recognition is reliable.
AI + YOU

Benchmark this project

Use the AI-on-the-edge-device prompt. Compare what different models deliver, then add what happened when you tried it.

Try this benchmark

View the project prompt

CAIBI adaptation · Prompt 1.0 · Project source

Prompt for AI-on-the-edge-device

CAIBI project benchmark: AI-on-the-edge-device
Prompt version 1.0

Help me build AI-on-the-edge-device using the existing creator project at https://github.com/jomjol/AI-on-the-edge-device. This is a CAIBI benchmark adaptation of the project build brief, not a claim about how the creator used AI.

Project context:
- An ESP32-CAM project that reads existing water, gas or electricity meters and publishes the readings digitally.
- Build context: ESP32 board with a supported camera; Integrated or external illumination; FAT formatted SD card; USB connection for initial flashing; Meter-specific mounting or 3D printed housing; Wi-Fi network for setup, updates and integrations.
- Where AI helps: Recognition, Configuration help, Troubleshooting.
- Physical work: Mount camera, Align, Configure network, Validate readings.
- Main limitation: Mounting, lighting and camera alignment are part of the system. Recognition software cannot compensate for a poor physical installation.

Read the source first. Use its published design, files, dimensions and parts as the baseline; do not redesign it unless a missing detail requires a clearly labelled proposal. If you cannot access a source or verify a revision, say so.

Provide one practical build package:
1. Design and files: explain the source design, link the original editable files, and identify missing files or modifications.
2. Parts and fabrication: give quantities, exact variants and compatibility, plus a cutting or printing list where applicable. Preserve source dimensions; flag anything unconfirmed.
3. UK and US purchasing: direct product links, checked dates, stock, pack sizes, taxes, delivery and regional costs. Mark unverifiable details as unknown; do not invent links, prices or availability.
4. Assembly: an ordered guide covering fabrication, fitting, wiring or software as applicable, setup, checks and troubleshooting.
5. Cost and time: separate parts, fabrication, delivery, hands-on time and machine time, with assumptions and exclusions.
6. Function and ease: a test plan for the project's stated purpose, likely difficulties, required skills and any corrections or limitations.

Keep source facts, your proposals and items needing physical verification distinct. An estimate is not a measured build result. Do not claim to have built or tested hardware. Include failed attempts and omissions. If a section does not apply, say why.

Use the source as accessed during this run and record its revision or access date. The prompt is fixed; the linked source can change.

Open and copy prompt →

Benchmark results 0

No reviewed results yet for AI-on-the-edge-device. Add the first result →

AI IN THIS PROJECT

Get AI help on this build

CAIBI · AI CAN HELP

AI can help with

Recognition
Configuration help
Troubleshooting

CAIBI · AI CAN'T DO

You still need to

Mount the camera rigidly in front of the real meter
Create stable lighting and remove glare from the physical installation
Choose and adjust the final camera angle and distance
Verify that the recognised reading matches the actual meter over time

Original source AI use and prompts

SOURCEThe original project does not publish an AI prompt

The prompts below are CAIBI helpers for this build. They are not presented as instructions from the creator.

CAIBI STARTING PROMPTS

Copy a prompt for this build

Each prompt tells AI to keep the original project source as the reference and to flag anything you must verify on the real build.

Ways to take it further

AFTER THE FIRST BUILD

Ways to take it further

Water meter readerGas meter readerElectricity meter readerHome Assistant integrationMQTT outputInfluxDB logging

More routes and resources

OTHER WAYS TO GET THERE

You do not have to follow one route.

ORIGINAL BUILDFollow the creator route

Use the published files and instructions as the baseline.

ALTERNATIVE PROJECTVerhoBot

A compact ESP32 curtain robot built to automate an existing curtain rail without ropes or pulleys.

REVIEWED RESOURCES

Useful places beyond the main source

DOCSAI-on-the-edge-device documentation

Current setup, compatibility and installation documentation.

SOURCE · CAIBI · BUILDERS

What supports the verdict

SOURCEjomjol

Source available with active documentation and releases

Open source ↗

CAIBIBUILD IT

It adds digital reporting to a meter that already works. The value is in avoiding replacement hardware and extracting more use from the equipment you already have.

Evidence: Strong

BUILDERSNo published reports yet

Actual time, cost, changes and outcomes appear here after submitted builds are reviewed.

Be the first builder to report ↓

Leave a build review

HAVE YOU MADE THIS?

Leave a build review.

First-hand evidence improves the assessment

What we want to know

What happened when you actually tried to make it.

Did it work?
Actual time and cost
What did you change?
What was harder than the source suggested?
Would you build it again?