OCR Acuracy

[packeTracer]

I am having some inconsistent results with the ocr option. It is accurate most of the time but sometimes a 6 will be a 3, an 8 a 6, and so forth.

There is a scalefactor option we have adjusted for some users and aren't sure if it is helping yet. Has this helped in practice for people in the past? If so what value helps? Any other suggestions to minimize errors for the OCR? We are dialing phone numbers which dialing the wrong number is not ideal.

[Dorian (MJT support)]

The higher the number you use, the more accurate it would be. But it would also use more resources. It is also dependent on the font/colour/background.

OCR_SCALEFACTOR 2 is the default. I generally see people reporting a good success rate with 4.

Code: Select all

Let>OCR_SCALEFACTOR=4

I suggest starting with 3, then 4, and so on.

I also suggest experimenting to see if the Find Object Wizard can see the phone numbers, as you may not need to use OCR in the first place. Or if it's in a browser, the Chrome/Edge functions.