Using Computer Vision (CV) to perform stock take for Part component count.

[swdee]

So one annoying thing about keeping accurate stock levels for Parts is having to count how many 0402/0603's etc are left on a cut piece of Tape. I can't be bothered with that so don't really record stock changes, I just visually look at the tape and estimate how many I have left and if its low then restock.

I was thinking that we could use the Scanner/Camera on phone to capture an image of the Tape and that could be run through a Computer Vision (CV) model (using OpenCV) to count the number of components (resistors/capacitors etc) that are left.

I put together a very quick test from a photo here to count 44 components.

Whilst a quick prototype script is possible, usually when it comes to CV, refining it and keeping the confidence/accuracy up requires quite some fine tuning.

So I am curious what other people do to manage stock levels? Is it a burdensome task you hate? What do you think of the idea of using CV to do the counting?

[swdee]

In terms of technical issues, there is some OpenCV support for PHP, but its quite out of date.

https://github.com/php-opencv/php-opencv - Archived, last updated 3 years ago.
https://github.com/hihozhou/php-opencv - No longer developed, last updated 6 years ago.

I prefer to keep a project using the same language (PHP) but it doesn't look practical to do that. One obvious choice is Python but that is a very ugly eco-system, however Go could be used, perhaps with a standalone backend with REST interface the PartDB PHP code could query. Any thoughts on that?