meter-reader

meter (water, gas, electricity) reader from image to digits

Goals :

• Read meter (water, gas, electricity)
• Hardware independant (works with any camera that can produce a fixed image)
• Keep simple

Process :

• Get image from any camera (local or http)
• Setup a python compatible apps to process the image (docker, flask or other python compatible apps)
• [onetime] set regions to process
• Produce digits from image

Solutions 1 - python only : (tested on Python 3.12.3)

pip install -r requirements.txt

python draw_regions.py

input image source and draw digits regions from left to right

python meter_reading.py

optional arguments : --image_source http://192.168.1.113/img_tmp/alg.jpg --model model.tflite --regions regions.txt --no-gui --no-output-image

For testing all models : --test-all-models --expected_result 177663

Solutions 2 - docker/webapp - flask_meter_reader :

pip install -r requirements.txt

python app.py

*- http://127.0.0.1:5000

• [onetime]Go to Draw Regions> input you image (http or local) and set regions

• Return to index and process you image (api is available)

Todos:

• make a hassio component (not actually possible cause nor tflite-runtime nor tensorflow supports python 3.13). Perhaps can hassio use venv ? The actual solution use the docker image
• make a scene text recognition model to recognise all digits in one shot. (actually investigating paddleOcr, OpenOcr, Parsec)

Sources and inspiration:

• The idea : https://github.com/jomjol/AI-on-the-edge-device/
• The original AI model : https://github.com/haverland/Tenth-of-step-of-a-meter-digit/tree/master/output
• New AI models : https://github.com/nliaudat/digit_recognizer

Licence:

• Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC-BY-NC-SA)
• No commercial use
• The AI model from haverland is under Apache Licence