Camera OCR / Sorting / AI learning

[alpine]

@gidesa
I'm not being told anything I didn't know. Please, pay more attention to my quote on markov in reply #13:

In general I don't use SIFT, though I experimented with Hough. Simple blob and edges and some pretty simple highschool math go a long way.

[marcov]

In general I don't use SIFT, though I experimented with Hough. Simple blob and edges and some pretty simple highschool math go a long way.

I'll be glad to hear a bit more details about those alternative (to the NN) OCR methods.  I'm using Tesseract OCR for a long time and while generally happy with it, there is some occurring faults specific for the (so called) AI - hallucinations and overfitting mainly. Thus I'm trying to learn about some different methods for doing it, preferably with a lesser I in the name.

Note the remark was from traditional machine vision viewpoint, not necessarily OCR. I'm not an OCR expert, I work in general machine vision.

I don't know if I said it in this thread, but the application that I have is for randomly orientated 7 line LCD like numeric fonts (but then etched into glass or ceramic material) .  You could consider that fitting into the "OCR font" class.

AFAIK there is some fuzzy logic approach to the recognition, but the appnotes I have read are of limited application, mainly for OCR tailored fonts.
The eigenvalues mentioned by Mark  are also an interesting approach, but I suspect there is also a "self-learning" part in that, so and the big I included.

Thanks in advance for any little idea you are willing to share.

Hough would be the second choice for my application, but that is because such fonts only have straight lines.

[gidesa]

@Alpine
Indeed I read and answer to this:

I'll be glad to hear a bit more details about those alternative (to the NN) OCR methods.

All algorithms, that I have listed, are alternative to Neural Networks. Although statistics algorithms, too. Maybe you don't want statistics algorithms. Maybe you know that algorithms. It's not clear from your message.

EDIT: perhaps you are disappointed for the phrase "YOU ... etc.". Sorry, that was a joke on generic "you", comprising me, too. Not a "you, Alpine".

[LV]

Maybe something here will be useful.

Kolmogorov-Arnold Networks (KANs) are promising alternatives of Multi-Layer Perceptrons (MLPs).

https://www.researchgate.net/publication/381125655_Kolmogorov-Arnold_Network_for_Satellite_Image_Classification_in_Remote_Sensing

https://github.com/KindXiaoming/pykan?tab=readme-ov-file