ssocr VS AI-on-the-edge-device

I get this kind of picture which is a bit shitty because of the low-contrast display and non ideal placement of the flash, but is actually good enough to OCR. The AI-on-the-edge project didn't work well for me because it takes ages to process one photo and I want better precision than that, so I rigged up a script that uses imagemagick to rotate, crop and threshold the image to something like that and then ssocr (github) does the job of actually recognizing the numbers.

This is what I am using for this exact purpose: https://github.com/jomjol/AI-on-the-edge-device

Anyway, in case decoding the radio traffic gets you nowhere, this works reasonably well.

- https://jomjol.github.io/AI-on-the-edge-device-docs/

I was looking for a way to collect readings of our water clock and stumbled upon this (https://github.com/jomjol/AI-on-the-edge-device) interesting hw/sw project to aquire data from old measuring devices (water, power, gas,...) with tflite and ocr.

I think this would be better https://github.com/jomjol/AI-on-the-edge-device. It's esp camera with ai detection for water meters, I use it also as monitoring for electricity.. the new firmware has HA support over mqtt. It's more diy but once you set it up it works like a charm.