Welcome to meter-reader Discussions!

[nliaudat]

👋 Welcome!

We’re using Discussions as a place to connect with other members of our community. We hope that you:

• Ask questions you’re wondering about.
• Share ideas.
• Engage with other community members.
• Welcome others and are open-minded. Remember that this is a community we
  build together 💪.

To get started, comment below with an introduction of yourself and tell us about what you do with this community.

[Fywqdk]

I think this is a great idea!

I have been having some problems with AI-on-the-edge, both related to a lot of wrong readings (which is probably an issue with light, which I will try to fix) and location (one of my meters is outside in a well. I can get wifi signal, but have no power, so battery and/or solar is needed.

Consequently I want to reduce the complexity on my various devices so I have to do minimal maintenance there. On the other hand having a docker container proces the images with OpenCV and python will allow for faster, and hopefully more stable data processing, while significantly lowering power consumption on the units at the meters.

Looking forward to seeing where this goes and maybe even contributing, though it is quite long since I last time looked at OpenCV.

[nliaudat]

Hi,
Glad I am not alone :)

OpenCV is only used to resize the image to model fit and output the final result. (It could be discarded).
The main process involves tensoflow or tflite-runtime (smaller).

For the hardware part, I use an esp-cam with esphome firmware.
https://github.com/nliaudat/meter-reader/tree/main/esphome

Actually, the led light is managed by my hassio.

There is still job to make a functionnal project
Regards

[Fywqdk]

Arh ok! Makes sense.

I am currently working on an esphome solution as well for hardware.

Regarding the Todo in the readme, and making it into a functional project: Would it make sense to make it into an addon rather than an integration? That way it can run any desired python version in its own container, including whichever is needed for the AI-packages. The addon should basically just be the flask webapp I guess? Then that could feed into a small integration or publish the result by MQTT or something?

[nliaudat]

I have never made hassio addon nor integration.
That's why it's a far project for me.

I am working on an AI model to recognize meter value from images in one shot (Scene Text Recognition).

I mean developping a cutom firmware is very hard and useless. Esphome has all functionnalities needed apart from the camera integration which need to be adapted for snapshot mode instead of allway capturing mode (for deepsleep integration)

[Fywqdk]

Same for me unfortunately. I will keep poking around with this but I am nowhere near competent enough to take lead on neither addon nor integration.